(navigation image)
Home American Libraries | Canadian Libraries | Universal Library | Community Texts | Project Gutenberg | Biodiversity Heritage Library | Children's Library | Advanced Microdevices Manuals | Linear Circuits Manuals | Supertex Manuals | Sundry Manuals | Echelon Manuals | RCA Manuals | National Semiconductor Manuals | Hewlett Packard Manuals | Signetics Manuals | Fluke Manuals | Datel Manuals | Intersil Manuals | Zilog Manuals | Maxim Manuals | Dallas Semiconductor Manuals | Temperature Manuals | SGS Manuals | Quantum Electronics Manuals | STDBus Manuals | Texas Instruments Manuals | IBM Microsoft Manuals | Grammar Analysis | Harris Manuals | Arrow Manuals | Monolithic Memories Manuals | Intel Manuals | Fault Tolerance Manuals | Johns Hopkins University Commencement | PHOIBLE Online | International Rectifier Manuals | Rectifiers scrs Triacs Manuals | Standard Microsystems Manuals | Additional Collections | Control PID Fuzzy Logic Manuals | Densitron Manuals | Philips Manuals | The Andhra Pradesh Legislative Assembly Debates | Linear Technologies Manuals | Cermetek Manuals | Miscellaneous Manuals | Hitachi Manuals | The Video Box | Communication Manuals | Scenix Manuals | Motorola Manuals | Agilent Manuals
Search: Advanced Search
Anonymous User (login or join us)
Upload
See other formats

Full text of "Papers on bacterial viruses"

coo 



0C60 



BACTERIAL 
VIRUSES 

Selected \hy 

GUNTHER S. STENT 



PAPERS ON BACTERIAL VIRUSES 



(5R 
^60 



PAPERS ON 



BACTERIAL VIRUSES 



Selected by 

GUNTHER S. STENT 

University of California, Berkeley 




Boston ^^^^^§y Toronto 
LITTLE, BROWN AND COMPANY 



COPYRIGHT © 1960, LITTLE, BROWN AND COMPANY, INC. 

ALL RIGHTS RESERVED. NO PART OF THIS BOOK 
MAY BE REPRODUCED IN ANY FORM WITHOUT 
PERMISSION IN WRITING FROM THE PUBLISHER. 

LIBRARY OF CONGRESS CATALOG CARD NO. 60-53146 



Published simultaneously in Canada 
by Little, Brown & Company {Canada) Limited 

PRINTED IN THE UNITED STATES OF AMERICA 



FOREWORD 



In 1951, the collection Papers in Microbial Genetics; Bacteria and Bacterial 
Viruses appeared under the editorship of Joshua Lederberg. That collection 
has been an indispensable aid to us and many others in the teaching of micro- 
bial genetics. It has enabled every student to read a cross-section of the 
original literature and thus become directly acquainted with the basic experi- 
mental data that often outlive the conclusions based upon them. 

The Lederberg collection, however, appeared just at the eve of great dis- 
coveries which, since that time, have enormously advanced our understanding 
of the structure and function of the hereditary substance of bacteria and their 
viruses. The transduction of bacterial genes by bacteriophages, the polarity of 
bacterial conjugation, the role of DNA as the germinal substance of bacterial 
viruses, the structure of DNA itself — all of these were discovered within the 
two-year period following the publication of Lederberg's collection. New 
techniques have also greatly changed the course and nature of research in 
microbial genetics; in particular, the use of isotopic tracers in bacterial virus 
research and the development of recombination tests of high resolving power 
for the analysis of genetic fine structure have brought microbial genetics to the 
molecular level. 

Hence it seemed urgent that a new, more up to date, collection should now 
be made available; we have decided to assume this task, and hope that the 
present book and its companion volume, Papers on Bacterial Genetics, will 
answer the needs of at least the immediate future. The extent of the changes 
that have taken place is indicated by the fact that of the more than fifty papers 
we have selected for the present two volumes, only five have been retained 
from the twenty papers of Lederberg's collection. 

In addition to the reprinted papers, each volume contains an introductory 
text with its own bibliography, in order to compensate for the necessarily in- 
complete and arbitrary nature of the selections. The selections themselves were 
made for the sole purpose of providing students with a cross-section of the 
literature, at a reasonable cost; in the words of Professor Lederberg "no 
apologies need be offered for a selection which must be largely arbitrary." 

In conclusion, we would like to express our deep appreciation to the copy- 
right owners and the authors who permitted us to reprint these papers, as well 
as to those who have supplied us with rare reprints from which some of the 
copy has been made. 

GuNTHER S. Stent 
Edward A. Adelberg 
Berkeley, California, 1960 



CONTENTS 



Introduction, by Gunther S. Stent ix 

Discussion on the Bacteriophage (Bacteriolysin). I. The Nature of 
Bacteriophage, by F. D'Herelle; II. The Bacteriophage: The 
Breaking Down of Bacteria by Associated Filter-passing Lysins, 
by F. W. Twort; III. Concerning the Theories of the So-called 
"Bacteriophage," by J. Bordet; IV. Andre Gratia 3 

Adsorption of Bacteriophages to Homologous Bacteria. II. Quantitative 
Investigations of Adsorption Velocity and Saturation. Estimation 
of the Particle Size of the Bacteriophage, by M. Schlesinger 26 

The Growth of the Bacteriophage, by Emory L. Ellis and Max DELBRtJCK 37 

The Growth of Bacteriophage and Lysis of the Host, by M. Delbrijck 57 

The Intracellular Growth of Bacteriophages. L Liberation of 

Intracellular Bacteriophage T4 by Premature Lysis with Another 

Phage or with Cyanide, by A. H. Doermann 75 

Independent Functions of Viral Protein and Nucleic Acid in Growth of 

Bacteriophage, by A. D. Hershey and Martha Chase 87 

Nucleic Acid Transfer from Parental to Progeny Bacteriophage, by 

J. D. Watson and O. Maal0e 105 

The Synthesis of Bacterial Viruses; Origin of Phosphorus Found in 
Deoxyribonucleic Acids of T2 and T4 Bacteriophages, by 
Seymour S. Cohen 116 

Enzymatic Synthesis of Deoxyribonucleic Acid. VI. Influence of 
Bacteriophage T2 on the Synthetic Pathway in Host Cells, by 
Arthur Kornberg, Steven B. Zimmerman, S. R. Kornberg, 
and John Josse 125 

The Frequency Distribution of Spontaneous Bacteriophage Mutants as 
Evidence for the Exponential Rate of Phage Reproduction, by 
S. E. LuRiA 139 

Genetic Recombination between Host-Range and Plaque-Type Mutants 
of Bacteriophage in Single Bacterial Cells, by A. D. Hershey and 
Raquel Rotman 151 

Genetic Recombination and Heterozygosis in Bacteriophage, by A. D. 

Hershey and Martha Chase . 179 



vn 



viii Contents 

The Structure of DNA, by J. D. Watson and F. H. C. Crick 193 

Fine Structure of a Genetic Region in Bacteriophage, by Seymour 

Benzer 209 

Induction of Specific Mutations with 5-Bromouracil, by Seymour 

Benzer and Ernst Freese 220 

Experiments on Photoreactivation of Bacteriophages Inactivated with 

Ultraviolet Radiation, by R. Dulbecco 228 

Genetic Recombinations Leading to Production of Active Bacteriophage 
from Ultraviolet Inactivated Bacteriophage Particles, by S. E. 
LuRiA and R. Dulbecco 247 

Inactivation of Bacteriophages by Decay of Incorporated Radioactive 

Phosphorus, by Gunther S. Stent and Clarence R. Fuerst 280 

Resistance to Ultraviolet Light as an Index to the Reproduction of 

Bacteriophage, by S. Benzer 298 

Investigations on a Lysogenic Bacillus Megaterium, by Andre Lwoff 

and Antoinette Gutmann 312 

Induction of Bacteriophage Lysis of an Entire Population of Lysogenic 
Bacteria, by Andre Lwoff, Louis Siminovitch and Niels 
Kjeldgaard 332 

Lysogeny and Genetic Recombination in Escherichia coli K12, by 

Elie L. Wollman and Franqois Jacob 334 

Spontaneous Induction of the Development of Bacteriophage \ in 

Genetic Recombination of Escherichia coli K12, by FRANgois Jacob 

and Elie Wollman 336 

Transduction of Lysogeny in Escherichia coli, by FRANgois Jacob 339 

Recombination between Related Temperate Bacteriophages and the 
Genetic Control of Immunity and Prophage Locahzation, by 
A. D. Kaiser and F. Jacob 353 



Pagination of ike papers in this collection is indicated by the boldface 

number centered at the bottom of the page. Other page numbers appearing 

on certain papers refer to original publication. 



INTRODUCTION 



Bacterial viruses were discovered in 1915, when F. W. Twort (149) isolated 
a filtrable virus which produces a "glassy transformation" of micrococcal 
colonies during their growth on an agar surface. Twort's paper remained 
relatively unnoticed, until two years later F. d'Herelle (64) published his own 
observations on a filtrable agent, the "bacteriophage," capable of serially trans- 
missible lysis of growing cultures of enteric bacilli. D'Herelle's announcement 
caused an immediate sensation among medical bacteriologists, since the 
bactericidal properties of the bacteriophage oflFered promise of a generalized 
prophylaxis and therapy of bacterial diseases. Within two or three years of 
the publication of his first paper, d'Herelle had carried out many incisive 
experiments which allowed him to recognize the essential aspects of these 
bacterial viruses. 

Twort and d'Herelle did not remain the only bacteriophage workers for 
very long, and the study of bacterial viruses rapidly became so popular that 
most of the leading bacteriologists of the decade following the first World 
War tried their hand at it. This soon led to a number of violent controversies 
concerning the nature and mode of action of bacteriophages. Some of these 
controversies were aired at a discussion on the bacteriophage organized by 
the British Medical Association at its Glasgow meeting of 1922. The presenta- 
tions which d'Herelle, Twort, Bordet, and Gratia prepared for this meeting 
form the first paper of this collection. In his discussion, d'Herelle demonstrates 
the self-reproducing, or viral, character of the bacteriophage, a view for which, 
as is evident from the remarks of Bordet and of Gratia, he had been under 
attack. Gratia's discussion, however, rectifies two errors of d'Herelle: the claim 
that the phenomena discovered by Twort and by d'Herelle are fundamentally 
difiFerent and the assertion that all bacteriophages represent the same antigen. 
In 1926, d'Herelle (65) thus summarized his earlier findings on the multi- 
plication of bacterial viruses: "The first act of bacteriophagy consists in the 
approach of the bacteriophage corpuscle toward the bacteria, then in the 
fixation of the corpuscle to the latter . . . The bacteriophage corpuscle pene- 
trates into the interior of the bacterial cell. When, as a result of its faculty of 
multiplication, the bacteriophage corpuscle which has penetrated into the 
bacterium fonns a colony of a number of elements, the bacterium ruptures 
suddenly, liberating into the medium young corpuscles which are then ready 
to continue the action." 

The first workers to study in some detail the initial step in bacterial virus 



IX 



X Introduction 

growth, i.e. the fixation, or adsorption, of the bacteriophage to its bacterial 
host cell, were Krueger (91) and Schlesinger (129, 130). In his paper (130), 
the second of this collection, Schlesinger demonstrates that phage adsorption 
usually is an irreversible process which follows the kinetics expected from a 
two-body collision model involving freely diffusing virus particles and bacterial 
cells. Schlesinger 's work, which probably represents the first rigorous applica- 
tion of physicochemical principles to the study of bacterial viruses, was 
extended by Delbriick (39), who showed that the physiological state of the 
bacteria affects the rate of adsorption, and by Garen and Puck (59), who 
demonstrated clearly that, under certain conditions, a reversible union between 
phage and bacterium can take place. Contrary to the predictions of the simple 
two-body collision model of Schlesinger, however, it turned out that the rate 
of phage adsorption reaches a maximum at high bacterial concentrations and is 
strongly temperature-dependent. Hence, it appears that the irreversible fixa- 
tion of the bacteriophage is a two-step process, involving at least one further, 
temperature-sensitive, step in addition to the collision of virus and host cell 
(146). Subsequent electron-optical observations by T. F. Anderson revealed 
that the organ of adsorption is the phage tail (5), in particular, that it is the 
thin tail fibers (155) which are the structures undergoing the stereospecific 
fixation reaction with the phage receptors on the bacterial surface. (Reviews: 
58, 148, 73, 125, 154. ) 

Convincing support for d'Herelle's conception that the infecting phage 
particle multiplies within the bacterium and that i*^s progeny are liberated 
upon lysis of the host cell was adduced in 1929 by Burnet (31), who showed 
that 20 to 100 viruses suddenly appear some 20 minutes after a bacterial 
suspension is infected with a single phage particle. The final demonstration, 
however, that a burst of progeny of the parent virus is liberated by each 
infected bacterial cell after a latent period was only provided in 1939 in the 
one step growth experiment of Ellis and Delbriick (52), whose paper is 
presented in this collection. In their publication, Ellis and Delbriick also 
describe for the first time the single burst experiment, which made possible 
the study of phage growth in individual infected bacteria, rather than in mass 
culture. The appearance of this paper marks the beginning of modern phage 
research. 

The interpretation of the phage-induced lysis of infected bacterial cultures 
was one of the violent controversies during the first 20 years of phage research. 
While d'Herelle (65) correctly thought that intracellular phage growth leads 
to lysis of the host cell and liberation of the virus progeny, Bordet and Ciuca 
(20, 22) maintained that the phage-induced dissolution of bacterial cultures 
is merely the consequence of a stimulation of lytic enzymes endogenous to the 
bacteria. Other workers, such as Bronfenbrenner (30), or Krueger and Nor- 
throp (92), thought that lysis of the bacteria is only a secondary phenomenon, 
which may or may not follow the growth of phage, and imagined that the 
bacteriophage can pass freely in and out of bacterial cells. Delbriick (40) 



Introduction xi 

finally showed, in his second paper of this collection, that these arguments 
were bedevilled by the circumstance that there exist not one but two com- 
pletely different processes by which bacteriophages can lyse susceptible bac- 
terial cells. One of these, lysis-from-without, represents an immediate dissolu- 
tion of bacteria, often encountered when the multiplicity of infection is much 
greater than one phage per bacterium (9, 123, 92, 124). Loss of the input 
phages, rather than their multiplication, ensues from this form of lysis. Only 
the second of the lytic processes, lysis-frorn-within, is really the form of lysis 
properly connected with intracellular phage multiplication, and its onset 
signals the end of the latent period. 

The one-step growth experiment demonstrated clearly the nature and 
kinetics of the process by which bacterial viruses multiply within cultures of 
susceptible bacteria. It thus brought into focus the question of fundamental 
biologic interest: what is taking place inside the infected cell during the latent 
period while the parental phage particle replicates itself several hundredfold? 
In order to study the kinetics of intracelltdar phage multiplication, Doermann 
(47) broke open phage-infected bacteria at various times during the latent 
period and assayed the infectivity of the material released by premature lysis. 
The result of this experiment, published in the paper included in this collec- 
tion, was that the infectivity associated with the original parental virus is lost 
at the outset of the reproductive process, since no infective particles whatso- 
ever can be found in any of the bacteria lysed within ten minutes after theix 
infection. After more than ten minutes have elapsed, however, ever-increasing 
numbers of infective progeny viruses make their intracellular appearance, 
until the final crop of progeny has been attained which would have been 
released by spontaneous lysis-from-within at the end of the normal latent 
period. The stage of intracellular bacterial virus growth during which the in- 
fected host cell contains no material capable of infecting another bacterium is 
the eclipse (107). Subsequent studies showed that the actual multiplication of 
the infecting virus takes place during the eclipse, i.e. that the phage multiplies 
in a non-infectious form, the vegetative phage (48). 

Schlesinger was also the first to purify a bacterial virus, a feat which he 
accomplished by high-speed centrifugation of phage lysates ( 131 ) . Chemical 
analysis of the purified virus revealed that it consists of approximately equal 
proportions of protein and deoxyribonucleic acid (DNA) (132). Later studies 
by Anderson (3, 4) and by Herriott (66) showed that the viral DNA resides 
within a proteinaceous head membrane, from which it can be released by 
osmotic shock. Hershey and Chase (75) then demonstrated that the two 
viral moieties, protein and DNA, have independent functions in the infection 
process. For, as can be seen in their first paper included in this collection, 
Hershey and Chase found that practically all of the viral protein remains at 
the surface of the infected cell, and that it is mainly the viral DNA which 
enters the bacterium at the outset of intracellular phage growth. The bulk of 
the phage protein appears to be relieved of any further function in the intra- 



xii Introduction 

cellular reproductive process after the proteinaceous tail has attached the virus 
particle to the bacterial surface and the DNA has safely entered the interior 
of the host cell. This historic discovery showed that it must be the viral DNA 
that is the carrier of the hereditary continuity, i.e. the germinal substance of 
the extracellular, resting phage. The release of the DNA from its protein 
envelope at the very moment of infection also accounts for the existence of the 
eclipse period at the early stages of intracellular virus development. For 
having been divested of its attachment and injection organs, the DNA of the 
infecting phage naturally is unable to gain entrance into any further bacterial 
cells to which it may be presented in the infectivity test. 

What happens to the viral DNA after its injection into the host cell? In 
1950 Putnam and Kozloff ( 126 ) devised an experiment directed toward the 
question of whether any of the atoms of the parental DNA ultimately reappear 
among the progeny viruses. In this "transfer experiment," bacteria are infected 
with phage particles whose DNA is isotopically labeled, and the phage yield 
issuing from such infected cells assayed for its content of parental isotope. The 
outcome of Putnam and KozlofiF's transfer experiment was that about half of 
the atoms of the parental DNA were found to be transferred to the progeny. 
This work was confirmed and extended with improved experimental techniques 
by Watson and Maal0e (120, 152), one of whose publications is included in this 
collection. In view of the inference that it is the DNA of the virus which 
carries the genetic continuity into the host cell, it seemed likely that an under- 
standing of the mechanism of transfer of DNA atoms from parent to offspring 
might afford valuable insight into the nature of the reproductive process. 
Further investigations have revealed that the parental DNA complement of a 
single parental virus is not transferred intact to a single progeny virus, but that 
the molecular patrimony is dispersed over several offspring phages (76, 143, 
101, 145). 

Some theories of the nature of phage multiplication envisaged that there 
are present, within the normal host bacterium, bacteriophage precursors whose 
metamorphosis into mature bacteriophages is merely triggered by the infecting 
phage particle (93). This view was finally dispelled in 1948 by an experiment 
of Cohen (37), designed to determine the origin of the substance of the 
progeny phages. This work, reported in a paper of this collection, represents 
the first use of radioisotopes in the study of bacterial viruses. By exposing 
bacterial cultures to P*-, either only prior to or only subsequent to their infec- 
tion with phage, and analyzing the virus progeny for their relative content of 
radioisotope, Cohen could show that most of the phage DNA is synthesized 
from materials still in the growth medium at the moment of infection; hence, 
the phage particles cannot have been derived from pre-existing bacterial pre- 
cursors. The complete kinetics of assimilation of phage DNA phosphorus were 
studied subsequently by modifications of Cohen's original method, by either 
adding to or withdrawing from the growth medium of bacterial cultures P^- 
at various times before or after their infection. The results of this work led to 



Introduction xiii 

the idea that, prior to its incorporation into intact, infective progeny particles, 
the phage DNA exists in an intrabacterial phage precursor pool ( 144, 71 ) . 

Further insight into the process of phage multiphcation was gained by the 
discovery of a variety of "incomplete" phage structures which possess one or 
another of the properties of the virus without being endowed with the power 
of self -reproduction, the most complex of all its attributes. Thus, premature 
lysis of infected bacteria at late stages of the eclipse period liberates newly 
synthesized proteinaceous material already possessing some of the antigenic 
properties of the intact bacteriophage (119). The total amount of phage 
antigen finally liberated upon spontaneous lysis of the cells, furthermore, 
generally exceeds that incorporated into infective progeny (119, 32, 46). 
Electron-optical observations of such lysates, furthermore, reveal the presence 
of structures whose morphology bears some resemblance to the characteristic 
shape of mature bacterial virus particles (161, 63). Prominent among these 
structures are the "doughnuts" which Levinthal and Fisher (103) found to 
appear during the eclipse and then to increase in number at about the same 
rate as the complete phage particles. Later studies have shown that the dough- 
nuts are, in fact, empty phage heads, and that the "maturation" of infective 
progeny at the end of the eclipse seems to represent the stable union of phage 
precursor DNA with phage precursor protein into structurally intact virus 
particles (88,89). 

Phage precursor protein and phage precursor DNA are not the only 
materials whose synthesis within the host cell is induced, or presided over, by 
the DNA of the infecting parental virus. For, at the outset of intracellular 
phage growth, the formation of some non-precursor proteins must proceed 
before replication of the viral DNA can begin. One of these "early" proteins 
was identified by Flaks and Cohen (53, 54) as the enzyme deoxycytidylate 
hydroxymethylase, essential for the synthesis of the specific components of the 
viral DNA, 5-hydroxymethylcytosine (160). Studies by Kornberg, Zimmer- 
man, Kornberg, and Josse (90), presented in a paper of this collection, 
revealed the phage-induced formation of four further enzymes, all of which 
are demonstrably involved in the synthesis and replication of the viral DNA. 
It is important to realize, therefore, that the phenotypic expression of the 
genetic substance of the phage is not confined solely to the construction of 
materials that find incorporation into the mature, infective progeny virus. 

In common with other organisms, bacterial viruses sport occasional 
hereditary variants, or mutants, in the course of their growth (34, 68, 133). 
These mutants can diflFer from their parents in a variety of characteristics, 
such as the type of plaque formed on agar seeded with sensitive indicator 
bacteria (67), the strains of bacteria which the phage can infect (105), or the 
physical or chemical properties of the virus particle (2, 43, 26). The mutation 
of the vegetative phage during its intracellular growth was used by Luria 
(108) to probe the nature of the self-duplication of the hereditary material of 
the infecting particle, as shown in a paper of this collection. In his experiment, 



xiv Introduction 

Luria examined the individual phage yield of many thousands of phage- 
infected bacterial cells and scored the single bursts for the presence of progeny 
viruses possessing a certain plaque-type mutant character. On the basis of 
the observed clonal frequency distribution of these mutants, Luria was able 
to infer that the replication of the hereditary material of the phage is geo- 
metric, i.e. that it proceeds by a number of successive cycles of self-duplica- 
tion, since other conceivable reproductive models, e.g. successive replications 
of the initial parental element or chain replication of the last element produced, 
would have led to mutant distributions quite diflFerent from that actually found. 
In 1946, a most important discovery was made independently by Delbriick 
and Bailey ( 44 ) and by Hershey ( 67 ) , who examined the genetic character of 
the phage yield issuing from bacterial cells infected with two related parent 
viruses differing from each other in two mutant factors. It was found that 
among the progeny of such mixed infection there appear virus offspring carry- 
ing one of the mutant factors of one and one of the mutant factors of the other 
of the two parents, demonstrating that bacterial viruses can undergo genetic 
recomhirmtion. The first detailed study of genetic recombination in phage 
was undertaken by Hershey and Rotman (77), their paper being included in 
our collection. This work showed that, on the basis of the frequency with 
which recombinant progeny for various mutated characters appear in such 
"crosses," it is possible to construct a genetic map of the phage on which the 
mutant loci can be arranged in a linear order. Hershey and Rotman also 
examined the frequency of complementary recombinant types in the yields of 
individual mixed infected bacteria and found that the formation of comple- 
mentary types does not seem to occur in a single event [Bresch (28) was able 
to establish this conclusion even more convincingly in a later study]. This 
fact led Hershey and Rotman to entertain the notion that recombination in 
phage might not be the consequence of a reciprocal exchange of preformed 
genetic structures, such as chromosomal recombination in higher forms, but 
that it might be an act incidental to the replication of the genetic material 
itself. This hypothesis, which came to be called "partial replicas'" (69), or 
"copy choice" ( 97 ) , now forms one of the basic concepts in the understanding 
of the molecular basis of self-duplication and genetic recombination. As more 
data concerning the process of genetic exchange in phage accumulated, it 
became evident that the theoretical analysis of a phage "cross" is a problem in 
population genetics. It was seen that within each mixedly infected bacterial 
cell, growth and recombination of the numerous vegetative phage replicas 
proceed concurrently. In 1953, Visconti and Delbriick (150) developed, there- 
fore, a theory which succeeded in explaining quantitatively the recombinant 
frequency observed in different phage crosses under various conditions. This 
theory assumes that replication and recombination of vegetative phages pro- 
ceeds in an intrabacterial pool, in which phages repeatedly mate pairwise 
and at random until lysis of the host cell, and from which pool the vegetative 
phages are withdrawn irreversibly for maturation into infective progeny 



Introduction xv 

phages. A concise statement of this theory can be found in Adams' book ( 1 ) , 
and a more generahzed formulation is presented in two later analyses of this 
problem (137,29). 

An important clue to the nature of the elementary recombinational event 
in phage was uncovered by Hershey and Chase (74), described in their second 
paper of this collection. They noted that among the progeny of mixed infec- 
tions about 2% of the particles are heterozygous, in that these individuals 
carry homologous loci, or alleles, from both parents of the cross. The heterozy- 
gosity is only partial, however, in that in any one heterozygote virus only a 
very limited segment of the genome is actually of biparental provenance, most 
of its loci being homozygous, or derived from only one or the other of the 
parents. The structure and behavior of these heterozygotes suggested to 
Hershey and Chase that the formation of heterozygotes and the formation of 
recombinants might be related processes. The nature of heterozygotes was 
considered further by Levin thai (100), who demonstrated that such viruses 
are recombinant for genetic loci on opposite sides of the limited region of 
heterozygosity. Levinthal then inferred that recombinant phages, in fact, arise 
through the formation of heterozygotes in the course of phage reproduction by 
the partial replica, or copy-choice, recombination mechanism. This inference 
found further support from Levinthal's calculation that the observed frequency 
of heterozygotes is great enough to explain the observed frequency of recom- 
binational events. 

Once the viral DNA had been identified as the germinal substance, it 
became possible to consider in actual chemical terms how the hereditary 
information is stored in the resting phage and how it is replicated in the 
vegetative phage. After deoxyribonucleic acid was discovered by Miescher 
in 1871, some 60 years of chemical study of this substance revealed that its 
building block is the nucleotide, composed of one molecule each of phosphoric 
acid, deoxyribose, and either adenine, guanine, thymine, or cytosine. More 
recently, it was established that DNA molecules are, in fact, polymers of very 
high molecular weight, each molecule containing more than 10* nucleotide 
units joined through phosphate diester bonds linking successive deoxyribose 
molecules (cf. 36). The actual molecular architecture of DNA was worked 
out finally by Watson and Crick (151), whose paper is presented here. Wat- 
son and Crick showed that the DNA molecule consists of two helically inter- 
twined polynucleotide chains laterally held together by a pair of hydrogen 
bonds between a complementary pair of purine and pyrimidine residues on 
opposite chains. The nature of the DNA molecule suggests that the only 
specific aspect which could distinguish one DNA macromolecule from another 
is the precise sequence of the four possible purine-pyrimidine base pairs along 
the complementary nucleotide chains, i.e. that the hereditary information is a 
message written into the DNA macromolecule in an alphabet containing four 
letters. This structure also suggested to Watson and Crick a mechanism by 
which the DNA molecule could replicate itself; for if the two complementary 



xvi Introduction 

polynucleotide chains separate and each parental chain acts as the template 
for the de novo synthesis of a complementary daughter chain, a pair of DNA 
molecules would be generated, each half-old, half-new, whose specific purine- 
pyrimidine base pair sequence is identical to that of the parent molecule. A 
genetic mutation, from this point of view, would then be a rare copy error in 
the replication process by which a nucleotide carrying an incorrect base is 
introduced into the replica nucleotide chain, thus producing a change in the 
genetic information. Even though a number of modifications of this replication 
scheme of Watson and Crick were subsequently proposed (45), later experi- 
ments have shown that in the replication of bacterial DNA the distribution of 
the atoms of the parental molecules appears to proceed by the semi-conserva- 
tive route (122), a central feature inherent in the Watson-Crick scheme. 

One of the first successful attempts to bridge the gap between chemistry 
and genetics was made by Benzer (12) in his first paper of this collection. 
Benzer discovered a method for scoring very rare recombinant viruses appear- 
ing in phage crosses between parents bearing extremely closely linked mutant 
loci. This allowed him to construct a jine structure map of a large collection 
of mutants situated in a very restricted region of the phage genome. In 
consequence of this work, the concept of the gene, traditionally regarded as 
the unit of recombination, mutation and function, became clarified. For 
Benzer showed that these three aspects of the genetic material are opera- 
tionally separable and hence cannot share a common unit. Translated into 
molecular terms, the unit of recombination appears to represent one, or a few, 
nucleotide pairs along the DNA molecule, whereas the unit of mutation can 
be of variable length, ranging from the alteration of a single nucleotide pair, 
in case of a point mutation, to long-span alterations of the phage genome, 
covering hundreds or thousands of nucleotide pairs. Finally, the unit of 
function, or cistron, assumed to determine the specific chemical structure of an 
enzyme protein, or more precisely, of a polypeptide chain, is of the order of 
1000 nucleotides in length (13). 

After Benzer had arranged his set of closely linked spontaneous phage 
mutants into a linear linkage map, it became obvious that there exists a great 
variability in mutability of different genetic sites within a single functional 
group, or cistron, since at some loci, or "hot spots," spontaneous mutations recur 
with much greater frequency than at other, nearby loci. This differential 
mutability of individual genetic sites very probably reflects the chemical 
structure of the hereditary molecule corresponding to each locus; e.g. the 
chance of making a spontaneous copy error at a given site in the course of 
viral DNA replication might depend on which particular sequence of purine 
and pyrimidine residues obtains there and which particular base substitution 
will produce the mutant genotype in question (13). Benzer and Freese (14), 
therefore, examined also distribution, or mutational spectrum, of mutants 
induced by the action of chemical mutagens, in particular by replacement of 
thymine by its analog 5-bromouracil in the viral DNA, which replacement, as 



Introduction xvii 

Litman and Pardee had found (104), is highly mutagenic in bacterial viruses. 
Benzer and Freese's investigation, presented in this collection, showed that 
the set of mutants induced by the action of 5-bromouracil is completely 
different from the set of spontaneous mutants in the same general region of 
the viral genome, demonstrating that "the mutagen does not merely enhance 
the over-all mutation rate, but acts at specific locations in the hereditary struc- 
ture." Mutational spectra of other chemical mutagens were subsequently 
established, and it turned out that each of these substances raises the proba- 
bility of mutation at a restricted number and individually characteristic set of 
sites (27, 57). Further insight into the chemical nature of the induced muta- 
tions was provided by studies that determined the connection between the 
induction of a mutation at a specific site by a given mutagen and the ability of 
the same, or of another mutagen, to revert this mutation to the original state. 
On the basis of these results, Freese proposed that there exist two basic types 
of point mutation in the viral genetic material: transversions, corresponding to 
the substitution of a purine by a pyrimidine residue, or vice versa, and transi- 
tions, corresponding to the replacement of one type of pyrimidine by the other 
or of one type of purine by the other ( 56 ) . 

Not long after the discovery of the bacteriophage it was found that ultra- 
violet light (UV) kills the virus particle (65), and since then, UV has been 
the inactivation agent whose effects have been most extensively studied ( 136, 
95, 110, 139). This work has shown that in addition to simply destroying the 
reproductive power, UV also produces a number of important physiological 
and genetic effects. The inactivated phages, furthermore, are by no means 
inert, being still capable of adsorbing to and killing bacteria, and of interfer- 
ing with the growth of other, unirradiated phages in the same host cell (111). 
Some of the lethal effects of UV, finally, are reversible under appropriate con- 
ditions. An important example of such reversibility is the existence of photo- 
reactivation, discovered by Dulbecco (49) in the work presented herein. Dul- 
becco found that viability is restored to UV-inactivated phages if bacteria 
infected with such "dead" particles are illuminated with visible light. Dul- 
becco's quantitative analysis of photoreactivation showed that a fraction of 
the UV lesions, the photoreactivable sector, is restored by a light-activated 
enzyme system of the bacterial host cell. Later work by Bowen (23, 24) 
revealed that photoreactivation consists of two steps: the first step is a dark 
reaction requiring no light, which generates the substances adsorbing and 
"activated" by the quanta of visible light for the second, actually reactivating 
step. Experiments by Lennox, Luria, and Benzer (99) suggested that photo- 
reactivation constitutes a direct reversal rather than a bypass mechanism of the 
primary ultraviolet damage, a conclusion that now seems certain since Good- 
gal, Rupert, and Herriot demonstrated the in vitro photoreactivation of UV- 
inactivated transforming DNA by illuminated bacterial extracts (61). 

Viability can also be restored to UV-inactivated phages if two or more 
"dead" particles, each unable to reproduce itself in solo, happen to infect 



xviii Introduction 

the same bacterial cell. This is the phenomenon of multiplicity reactivation, 
discovered by Luria in 1947 (106) and investigated in some detail by Lm-ia 
and Dulbecco (112), whose paper is included in this collection. The quan- 
titative results presented here seemed to bear out Luria's proposal that each 
inactivating UV lesion represents a lethal mutation in one of a certain number 
of genetic subunits of the phage and that multiplicity reactivation ensues from 
the genetic exchange of still undamaged units between the two irradiated 
parent viruses. In order to explain the very high frequency of reactivation, 
furthermore, it was assumed that phage growth occurs by the independent 
reproduction of each subunit, followed by reassembly of the units into com- 
plete phages. When Dulbecco (50) subsequently continued his multiplicity 
reactivation studies, he found that the results observed at very high UV doses 
are no longer compatible with the notion of independently multiplying sub- 
units. In any case, studies on genetic recombination in phage had in the mean- 
time indicated that the genetic material of the phage does not multiply in the 
form of independent subunits (150). More recently, however, modifications 
of the original hypothesis of multiplicity reactivation have been proposel by 
Baricelli ( 10 ) and by Harm ( 62 ) , which still retain that most essential element 
of Luria's hypothesis that reactivation proceeds by a mechanism of genetic 
exchange of undamaged parts and which lead to quantitative formulations in 
satisfactory agreement with the observed data. 

Another radiobiological method of inactivation bacterial viruses was dis- 
covered by Hershey, Kamen, Kennedy, and Gest (76), who showed that 
highly P^ --labeled bacteriophages lose their infectivity upon decay of radio- 
phosphorus atoms. From the kinetics and efficiency of this inactivation process, 
it could be inferred that the cause of death is the transmutation of phosphorus 
into sulfur atoms in the polynucleotide chains of the viral DNA, or the highly 
energetic nuclear recoil associated with this event. These studies were 
extended by Stent and Fuerst ( 141 ) , whose paper appears in this collection. 
They found that, although the efficiency of killing of one lethal hit per ten 
P^- disintegrations first observed by Hershey and his co-workers also obtains 
in a variety of different bacteriophage strains, the fraction of disintegrations 
that are lethal depends on the temperature at which decay is allowed to 
proceed. A mechanism for the decay inactivation process of the virus was 
suggested on the basis of these findings. It was proposed that the high pro- 
portion of nonlethal decays reflects the possibility that the physiological func- 
tion of the double-stranded DNA molecule is preserved even after radioactive 
decay has interrupted only one of its polynucleotide strands. The lethal decays, 
in contrast, are thought to be those that result by chance in a complete cut 
of both strands of the DNA double helix. The decay of incorporated P^- atoms 
has proven a very useful tool for the study of the structure, physiology, and 
genetics of bacterial viruses and bacteria. (Review: 142.) 

In the hope of measuring the extent to which the infecting parental virus 
has multiplied within the host cell during the eclipse period before the appear- 



Introduction xix 

ance of any mature progeny, Luria and Latarjet (113) irradiated phage- 
infected bacteria with UV at various stages of intracellular phage growth. 
They reasoned that if the UV sensitivity of the vegetative phage is equal to 
that of the free, extracellular virus, then the result of this irradiation experi- 
ment ought to be a family of multiple-hit survival curves from which the instan- 
taneous number of vegetative phages present at the time of irradiation could 
be inferred. The outcome of Luria and Latarjet's experiment was contrary to 
their expectation, however; instead of the anticipated multiple-hit survival 
curves, a family of straight lines of ever decreasing slope was observed, indica- 
tive of the fact that the intrinsic UV sensitivity of the vegetative phage is much 
less than that of the free virus. This is also evident from Benzer's (11) im- 
proved experimental design of the Luria-Latarjet technique, presented here as 
Benzer's third paper. The meaning of the great reduction in UV sensitivity 
of the vegetative phage has not yet found an entirely satisfactory explanation. 
On the one hand, as is evident from Benzer's report, some phages do not mani- 
fest this effect, so that UV irradiation of bacteria infected with such phages 
actually gives rise to the family of multiple-hit curves anticipated by Luria 
and Latarjet. On the other hand, the vegetative phage is also much more 
resistant to inactivation by decay of incorporated P'^'- atoms than the extra- 
cellular P^'-labeled virus (138). It seems likely, however, that the reduction in 
radiosensitivity of the vegetative phage reflects some important aspect of the 
function and replication of the viral DNA, and some of the possible interpre- 
tations have been discussed in several reviews (94, 136, 139, 142). In any 
caise, the method of Luria and Latarjet has found a number of valuable 
applications in the study of intracellular virus growth, not only with bacterio- 
phages but also with plant and animal viruses (134, 135, 51, 128). 

Within a few years of the discovery of the bacteriophage, lysogenic 
bacterial strains were found which appear to "carry" bacteriophages, in the 
sense that phage particles are always present in the culture fluid of such 
strains (21, 60). It was soon realized that this association of phage and 
bacteria cannot be of a casual nature, since it is impossible to permanently 
free lysogenic strains from the phage they carry by methods which ought to 
kill or remove the virus particles, such as heating, anti-phage serum neutraliza- 
tion, or single-colony purification (8, 18, 121). The nature and significance of 
lysogeny then remained a subject of intense controversy for about 30 years, 
some workers denying the existence of "true" lysogeny and others claiming 
that lysogeny disproves the whole notion that bacteriophagy involves an infec- 
tion of bacteria by virus particles. Nevertheless, a few bacteriologists, such as 
Burnet and McKie (35), and the elder Wollman (156), already envisaged 
that lysogeny represents an innate capacity of bacterial cells for phage produc- 
tion. In order to establish firmly some of the basic but controversial facts of 
lysogeny, Lwoff began a study of this phenomenon after the Second World 
War and in 1950 published the paper (116) presented here. In this work 
Lwoff and Gutmann demonstrate unequivocally that each bacterium of a lyso- 



XX Introduction 

genie strain harbors and maintains a noninfeetive strueture, the probacterio- 
phage or prophage, which endows the cell with the ability to give rise to infec- 
tive phage without further intervention of exogenous virus particles. The 
actual synthesis of infective phage, however, proceeds in only a small fraction 
of the cells of a growing culture of lysogenic bacteria, whose intracellular 
content of virus particles is liberated by lysis of the phage-producing indi- 
vidual. LwofiF also inferred from these experiments that the induction of 
phage development in a lysogenic cell is under control of external factors, 
and his subsequent investigations in collaboration with Siminovitch and 
Kjelgaard (118) showed that treatment with various agents, in particular 
irradiation with UV light, will indeed induce phage production and ultimately 
lysis of almost every cell of a culture of lysogenic bacteria. The first pre- 
liminary report of this finding (117) is included in this collection. After the 
publication of Lwoff's papers and reviews (114), the study of lysogeny not 
only flowered into a distinct branch of bacterial virus research but also became 
the bridge leading from the genetics of virus to that of host cell. In fact, the 
recognition of the existence and nature of the proviriis state engendered 
entirely new ideas concerning the origin, evolution, and biological function of 
viruses (79). (Reviews: 114,17,85.) 

What is the relationship of the prophage to the remainder of the lysogenic 
cell? The great stability of the lysogenic character implies that the prophage 
is transmitted to daughter cells at each bacterial division. This could happen 
in one of two ways: either the prophage represent? numerous autonomous 
structures, replicating in the bacterial cytoplasm in synchrony with the rest of 
the bacterium and being partitioned at random at each division over the 
daughter cytoplasms, or the prophage is integrated into the nuclear apparatus 
of the host cell and participates in the specific replication and segregation 
process which assures, nolens volens, that each daughter cell obtains one 
complete set of parental hereditary factors. The first of these alternatives soon 
appeared unlikely, when indirect estimates of the number of prophages 
revealed that each cell seems to carry only one or two prophages per bacterial 
nucleus (16, 81), and that a given type of prophage appears to saturate a 
limited number of sites on some bacterial structure (15). Positive indications 
that the prophage is integrated into the bacterial nucleus became available 
from experiments in bacterial conjugation, in which non-lysogenic bacteria 
were crossed with lysogenic bacteria, and a linkage of the lysogeny character 
with other known genetic factors of the cell inferred from the segregation 
pattern of the recombinants (96, 158, 7, 55). After the discovery of high- 
frequency-recombination (Hfr) bacterial strains and of the oriented transfer 
of the bacterial chromosome from donor to recipient cell, Wollman and Jacob 
( 159 ) could show very clearly in their first paper of this collection that at least 
one particular prophage has its specific location on the bacterial chromosome; 
their later work, furthermore, revealed that different prophages have different 
specific chromosomal sites (83, 84). At the same time, Jacob and Wollman 



Introduction xxi 

(82) also discovered the existence of zygotic induction, described here in their 
second paper. For if a chromosome fragment of a donor bacterium bearing a 
prophage enters a non-lysogenic recipient cell, then the prophage becomes 
induced, enters the vegetative state, leading to the production of infective 
progeny and lysis and loss of the bacterial zygote. This phenomenon accounted 
for the discrepancies that had been observed in earlier attempts to determine 
the chromosomal location of the prophage by bacterial conjugation experi- 
ments, from which diflFerent linkage relations could be inferred, depending 
upon which of the two parents of the cross carried the prophage (160). 

A completely independent confirmation of the specific location of the pro- 
phage on the bacterial chromosome was provided through transduction 
experiments by Jacob (78), whose paper is presented here. Transducing 
virus particles carry a small genetic segment of closely linked loci of a donor 
bacterium, i.e. the last host cell, into a recipient bacterium, i.e. the next host 
cell. In this way, recombinant bacteria can arise which have derived a very 
limited part of their genome from the donor cell (164, 163, 147, 98). Jacob 
thus showed that a transducing virus can carry the prophage of an entirely 
unrelated virus strain from a lysogenic donor into a non-lysogenic recipient 
bacterium, usually in association with the contiguous region of the donor 
chromosome. The chromosomal region, moreover, turned out to be the same 
in which the prophage had been already placed by bacterial conjugation 
studies. 

The presence of the prophage not only endows the bacterium with the 
capacity to produce phage but also confers upon the cell an immunity to 
infection by a homologous phage (8, 18, 121, 157). Such virus particles are 
usually adsorbed to immune lysogenic bacteria, but the particles neither 
multiply to give rise to infective progeny nor afiFect growth and division of the 
immune cells in any way ( 16, 15, 81 ) . The immune character is highly specific, 
in that a bacterium lysogenic for, and hence immune to, infection by one 
type of phage is not immune to infection by other viruses whose prophage 
the cell does not happen to carry. It became possible to study the genetic 
basis of immunity when Wollman and Jacob (84) discovered a number of 
related phage strains that undergo genetic recombination with one another 
but that differ in their immune specificity as well as in the locations of their 
prophages on the host linkage map. The final paper of our collection presents 
the work of Kaiser and Jacob (87) which established by means of crosses of 
these related phage strains that there is a definite segment of the viral genome, 
the C region, which determines the immune specificity of the phage. This 
same segment also controls the ability of the virus to establish itself as pro- 
phage and the locus at which the prophage is situated on the bacterial chromo- 
some. Later experiments by Jacob and Campbell (80) have shown that im- 
mimity derives from the presence of a repressor substance in the cytoplasm 
of the lysogenic cell. The specificity of both formation and action of this 
repressor appears to be determined by the C region of the phage genome. 



xxii Introduction 

Before closing this introduction, we must once more call attention to the 
necessarily very limited and quite arbitrarily restricted scope of this collection. 
Not only do the twenty-five papers presented make up a very small fraction of 
the important contributions to the study of bacterial viruses, but even the 
references listed on the following pages are by no means an adequate, or even 
representative, coverage of the relevant literature. Some impression of the 
extent of that literature can be gained from the astounding bacteriophage 
bibliography prepared by Raettig (127), which lists no fewer than 5655 
papers published between 1917 and 1956! The reader who wishes to acquire 
a more profound knowledge of this subject without working his way through 
these 5655 papers may do so by studying some of the books and reviews on 
bacterial viruses which have appeared over the years. This study might begin 
with Adams' book Bacteriophages (1), the most authoritative and compre- 
hensive treatment available at present, and with the more specialized chapters 
in Burnet and Stanley's treatise The Viruses (115, 58, 140, 102, 85, 136). More 
easily digestible works are Weidel's little book Virus ( 153 ) , a sprightly intro- 
duction to bacterial viruses, and Afinsen's The Molecular Basis of Evolution 
(6) which contains an account of the role of bacterial viruses in avant-garde 
molecular biology. A first-hand acquaintance with the evolution of bacterial 
virus research can be obtained by perusal of the reviews of Bronfenbrenner of 
1928 (30), Bordet of 1931 (19), Burnet of 1934 (33), Delbriick of 1942 and 
1946 (41, 42), Cohen of 1949 (38), Luria of 1950 (107) and Hershey of 1952, 
1953, 1956, and 1957 (69, 70, 72, 73). Finally, some other important reviews 
that treat special aspects of bacterial viruses are those of Luria (109), Brenner 
(25), and Jacob and Wollman (86, 79). 



BIBLIOGRAPHY 



1. Adams, M. H. "Bacteriophages," Interscience, New York, 1959. 

2. Adams, M. H., and Lark, G. Mutation to heat resistance in coliphage T5. /. 
Immunol, 64, 335 (1950). 

3. Anderson, T. F. The reactions of bacterial viruses with their host cells. Botan. Rev., 
15, 464 (1949). 

4. Anderson, T. F. Destruction of bacterial viruses b}^ osmotic shock. J . Appl. Phys., 
21, 70 (1950). 

5. Anderson, T. F. The morphology and osmotic properties of bacteriophage systems. 
Cold Spring Harbor Symp. Quant. Biol., 18, 197 (1953). 

6. Anfinsen, C. B. "The Molecular Basis of Evolution." John Wiley and Sons, New 
York, 1959. 

7. Appleyard, R. Segregation of lambda lysogenicity during bacterial recombination in 
Escherichia coli K12. Genetics, 39, 429(1954). 

8. Bail, O. Der Colistamm 88 von Gildemeister und Herzberg. Med. Klin. (Munich), 
21, 1271 (1925). 



Bibliography xxiii 

9. Bail, 0., and Matusmoto, T. Die Anhaufungsmoglichkeit von Bakteriophagen und 
Bakterien, Med. Klin. (Munich), 19, 1579 (1923). 

10. Baricelli, N. A. A "chromosomic" recombination theory for multiplicity reactivation 
in phages. Acta biotheoretica (Leiden), 11, 107 (1956). 

11. Benzer, S. Resistance to ultraviolet light as an index to the reproduction of bacteri- 
ophage. J. Baderiol, 63, 59 (1952). 

12. Benzer, S. Fine structure of a genetic region in bacteriophage. Proc. Natl. Acad. Sci. 
U.S., 4U 344 (1955). 

13. Benzer, S. The elementary units of heredity, in "The Chemical Basis of Heredity," 
W. D. McElroy and B. Glass, eds., p. 70. The Johns Hopkins Press, Baltimore, 1957. 

14. Benzer, S.. and Freese, E. Induction of specific mutations with 5-bromouracil. Proc. 
Natl. Acad. Sci. U.S., ^4, 112 (1958). 

15. Bertani, G. Lysogenic versus lytic cycle of phage multiplication. Cold Spring Harbor 
Symp. Quant. Biol, 18, 65 (1953). 

16. Bertani, G. Infections bact^riophagiques secondaires des bacteries lysogenes. Ann. 
Inst. Pasteur, 84. 273 (1953). 

17. Bertani, G. Lysogeny. Advances in Virxis Research, 5, 151 (1958). 

18. Bordet, J. Le probleme de I'autolyse microbienne transmissible ou du bacteriophage. 
Ann. Inst. Pasteur, 39,711 (1925). 

19. Bordet, J. The theories of the bacteriophage. Proc. Roy. Soc. London, B83, 398 
(1931). 

20. Bordet, J., and Ciuca, M. Exudats leucocytaires et autolyse microbienne transmis- 
sible. Compt. rend. Soc. Biol., 83, 1293 (1920). 

21. Bordet, J., and Ciuca, M. D^terminisme de I'autolyse microbienne transmissible. 
Compt. rend. Soc. Biol, 84, 276 (1921). 

22. Bordet, J., and Ciuca, M. Le bacteriophage de d'H^relle, sa production, et son 
interpretation. Compt. rend. Soc. Biol, 83, 1296 (1920). 

23. Bowen, G. H. Studies of ultraviolet irradiation phenomena — An approach to the 
problems of bacteriophage reproduction. Cold Spring Harbor Symp. Quant. Biol, 18, 
245 (1953). 

24. Bowen, G. H. Kinetic studies on the mechanism of photoreactivation of bacteri- 
ophage T2 inactivated by ultraviolet light. Ann. Inst. Pasteur, 84, 218 (1953). 

25. Brenner, S. Physiological aspects of bacteriophage genetics. Advances in Virus 
Research, 6, 137 (1959). 

26. Brenner, S., and Barnett, L. Genetic and chemical studies on the head protein of 
bacteriophages T2 and T4, in "Structure and Function of Genetic Elements." 
Brookhaven Symposia in Biology, No. 12, p. 86 (1959). 

27. Brenner, S., Benzer, S., and Barnett, L. Distribution of proflavin-induced mutations 
in the genetic fine structure. Nature, 182, 983 (1958). 

28. Bresch, C. Zum Paarungsmechanismus von Bakteriophagen. Z. Naturforsch., 10b, 
545 (1955). 

29 Bresch, C, and Starlinger, P. Zum Problem der genetischen Reko^nbination von 
Bakteriophagen. Z. Vererbungsl, 89, 459 (1958). 

30. Bronfenbrenner, J. J. In "Filtrable Viruses" (T. Rivers, ed.), p. 373. Williams and 
Wilkins, Baltimore, 1928. 



xxiv Introduction 

31. Burnet, F. M. A method for the study of bacteriophage multiplication in broth. 
Bnt. J. Expt. Path., 10, 109 (1929). 

32. Burnet, F. M. A specific soluble substance from bacteriophages. Brit. J. Exptl. 
Pathol., 14, 100 (1933). 

33. Burnet, F. M. The bacteriophages. Biol. Revs. Biol. Proc. Cambridge Phil. Soc, 9, 
332 (1934). 

34. Burnet, F. M., and Lush, D. Induced lysogenicity and mutation of bacteriophage 
within lysogenic bacteria. Australian J. Exptl. Biol. Med. Sci., I4, 27 (1936). 

35. Burnet, F. M., and McKie, M. Observations on a permanently lysogenic strain of B. 
enteridis gaertner. Austral. J. Exptl. Biol. Med. Sci., 6, 277 (1929). 

36. Chargaff, E., and Davidson, J. N. "The Nucleic Acids," Academic Press, New York, 
1955. 

37. Cohen, S. S. Synthesis of bacterial viruses; origin of phosphorus found in desoxyribo- 
nucleic acids of T2 and T4 bacteriophages. J. Biol. Chem., 174, 295 (1948). 

38. Cohen, S. S. Growth requirements of bacterial viruses. Bacteriol. Rev., 13, 1 (1949). 

39. Delbriick, M. Adsorption of bacteriophages under various physiological conditions 
of the host. /. Gen. Physiol., 28, 631 (1940). 

40. Delbriick, M. The growth of bacteriophage and lysis of the host. J . Gen. Physiol., 
23, 643 (1940). 

41. Delbriick, M. Bacterial viruses. Advances in Enzymol., 2, 1 (1942). 

42. Delbriick, M. Bacterial viruses or bacteriophages. Biol. Rev., 21, 30 (1946). 

43. Delbriick, M. Biochemical mutants of bacterial viruses. /. Bacteriol., 56, 1 (1948). 

44. Delbriick, M., and Bailey, W. T. Induced mutations in bacterial viruses. Cold 
Spring Harbor Symp. Quant. Biol., 11, 33 (1946). 

45. Delbriick, AI., and Stent, G. S. On the mechanism of DNA replication, in "The 
Chemical Basis of Heredity," W. D. McElroy and B. Glass, eds., p. 699. The Johns 
Hopkins Press, Baltimore, (1957). 

46. DeMars, R. I. The production of phage-related material when bacteriophage 
development is interrupted by proflavine. Virology, 1, 83 (1955). 

47. Doermann, A. H. The intracellular growth of bacteriophages. I. Liberation of 
intracellular bacteriophage T4 by premature lysis with another phage or with 
cyanide. J. Gen. Physiol, 35, 645 (1952). 

48. Doermann, A. H. The vegetative state in the life cycle of bacteriophage: Evidence 
for its occurrence and its genetic characterization. Cold Spring Harbor Symp. Quant. 
Biol, 18, 3 (1953). 

49. Dulbecco, R. Experiments on photoreactivation of bacteriophages inactivated with 
ultraviolet radiation. /. Bacteriol., 59, 329 (1950). 

50. Dulbecco, R. A critical test for the recombination theory of multiplicity reacti- 
vation. J. Bacteriol. 63, 199 (1952). 

51. Dulbecco, R., and Vogt, M. Biological properties of poliomyelitis virus as studied by 
the plaque technique. Annals New York Acad. Sci., 61, 790 (1955). 

52. Ellis, E. L., and Delbriick, M. The growth of bacteriophage. J. Gen. Physiol, 22, 
365 (1939). 

53. Flaks, J. G., and Cohen, S. S., Virus-induced acquisition of metabolic function. 
I. Enzymatic formation of 5-hydroxymethyldeoxycytidylate. J. Biol Chem., 234, 
1501 (1959). 



Bibliography xxv 

54. Flaks, J. G., Lichtenstein, J., and Cohen, S. S., Virus-induced acquisition of meta- 
bolic function. II. Studies on the origin of the deoxycy tidy late hydroxymethylase 
of bacteriophage-infected E. coli. J. Biol. Chem., 234, 1507 (1959). 

55. Fredericq, P. Localisation du prophage sur le chromosome. Son intervention dans 
le taux de segregation apparent des recombinants d' Escherichia coli K12. Compt. 
rend. Soc. Biol, 148, 1501 (1954). 

56. Freese, E. On the molecular explanation of spontaneous and induced mutations, in 
"Structure and Function of Genetic Elements." Brookhaven Syrnposia in Biology. 
No. 12, p. 63 (1959). 

57. Freese, E. The difference between spontaneous and base-analogue induced muta- 
tions of phage T4. Proc. Natl. Acad. Set., 45, 622 (1959). 

58. Garen, A., and Kozloff, L. M. The initiation of bacteriophage infection, in "The 
Viruses," F. M. Burnet and W. M. Stanley, eds., Vol. II, p. 204. Academic Press, 
New York, 1959. 

59. Garen, A., and Puck, T. T. The first two steps of the invasion of host cells by 
bacterial viruses. J. Exptl. Med. 94, 177 (1951). 

60. Gildemeister, E. Ueber das d'Herellesche Phanomen. Berlin, klin. Wochschr. 58, 
1355 (1921). 

61. Goodgal, S. H., Rupert, C. S., and Herriott, R. M. Photoreactivation of Hemophilus 
Influenzae transforming factor for streptomycin resistance by an extract of Escher- 
ichia coli B. in "The Chemical Basis of Heredity," W. D. McElroy and B. Glass, eds. 
p. 341. The Johns Hopkins Press, Baltimore, 1957. 

62. Harm, W. On the mechanism of multiplicity reactivation in bacteriophage. Vi- 
rology, 2, 559 (1956). 

63. Hercik, F. Some observations about the morphology of bacteriophage. Experientia, 
6, 64 (1950). 

64. d'H^relle, F. Sur un microbe invisible antagoniste des bacilles dysent^riques. 
Compt. rend. Acad. Sci., 165, 373, (1917). 

65. d'H^relle, F. "The Bacteriophage and its Behaviour." Williams and Wilkins, 
Baltimore, 1926, p. 135. 

"j6. Herriott, R. M. Nucleic acid-free T2 virus "ghosts" with specific biological action. 
/. BacterioL, 61, 752 (1951). 

67. Hershey, A. D. Mutation of bacteriophage with respect to type of plaque. Genetics, 
3L 620 (1946). 

68. Hershey, A. D. Spontaneous mutations in bacterial viruses. Cold Spring Harbor 
Symp. Quant. Biol., 11,67 (1946). 

69. Hershey, A. D. Reproduction of bacteriophage. Intern. Rev. Cytol., Vol I. (1952), 
p. 119. 

70. Hershey, A. D. Inheritance in bacteriophage. Advances in Genet., 5, 89 (1953). 

71. Hershey, A. D. Nucleic acid economy in bacteria infected with bacteriophage T2. 
II. Phage precursor nucleic acid. ./. Gen. Physiol., 37, 1 (1953). 

72. Hershey, A. D. Chemistry and viral growth, in "Currents in Biochemical Research," 
D. E. Green, ed. p. 1, Interscience, New York, 1956. 

73. Hershey, A. D. Bacteriophages as genetic and biochemical systems. Advances in 
Virus Research, 4, 25 (1957). 

74. Hershey, A. D., and Chase, M. Genetic recombination and heterozygosis in bacterio- 
phage. Cold Spring Harbor Symp. Quant. Biol., 16, 471 (1951). 



xxvi Introduction 

75. Horshoy, A. D., and Chase, M. Independent functions of viral protein and nucleic 
acid in growth of bacteriophage. J . Gen. Physiol., 36, 39 (1952). 

76. Hershey, A. D., Kamen, M. D., Kennedy, J. W., and Gest, H. The mortality of 
bacteriophage containing assimilated radioactive phosphorus. J. Gen. Physiol. 34, 
305 (1951). 

77. Hershey, A. D., and Rotman, R. Genetic recombination between host range and 
plaque-tvpe mutants of bacteriophage in single bacterial cells. Genetics, 34, 44 
(1949). 

78. Jacob, F. Transduction of lysogeny in Escherichia coli. Virology, 1, 207 (1955). 

79. Jacob, F. Genetic control of viral functions, in "The Harvey Lectures," Vol. 54, 
p. 1, Academic Press, New York, 1960. 

80. Jacob, F., and Campbell, A. Sur le systeme de repression assurant I'immunit^ chez 
les bacteries lysogenes. Conipt. rend. Acad. Sci., 248, 3219 (1959). 

81. Jacob, F., and Wollman, E. L. Induction of phage development in lysogenic bacteria. 
Cold Spring Harbor Sy7np. Quant. Biol., 18, 101, (1953). 

82. Jacob, F., and Wollman, E. L. Induction spontanea du development du bacterio- 
phage au cours de la recombinaison g^n^tique, chez Escherichia coli K12. Compt. 
rend., 239, 317 (1954). 

83. Jacob, F., and Wollman, E. L. Genetic aspects of lysogeny, in "The Chemical Basis 
of Heredity," W. D. McElroy and B. Glass, eds., p. 468. The Johns Hopkins Press, 
Baltimore, 1957. 

84. Jacob, F., and Wollman, E. L. Sur les processus de conjugaison et de recombinaison 
g^netique chez Escherichia coli. IV. Prophages inductibles et mesure des segments 
g^netiques transf^r^s au cours de la conjugaison. Annates Inst. Pasteur, 95, 497 

(1958). 

85. Jacob, F., and Wollman, E. L. Lysogeny, in "The Viruses," F. M. Burnet and 
W. M. Stanley, eds. Vol. II, p. 319, Academic Press, New York, 1959. 

86. Jacob, F., and Wollman, E. L. The relationship between the prophage and the 
bacterial chromosome in lysogenic bacteria, in "Recent Progress in Microbiology," 
G. Tuneval, ed. Almquist and Wiksell, Stockholm, 1959. 

87. Kaiser, A. D., and Jacob, F. Recombination between related temperate bacterio- 
phages and the genetic control of immunity and prophage localization. Virology, 4, 
509 (1957). 

88. Kellenberger, E., and S^chaud, J., Electron microscopical studies of phage multipli- 
cation. II. Production of phage-related structures during multiplication of phages 
T2 and T4. Virology, 3, 256 (1957). 

89. Kellenberger, E., and S^chaud, J., Electron microscopical studies of phage multipli- 
cation. IV. The establishment of the DNA pool of vegetative phages and the 
maturation of phage particles. Virology, 8, 478 (1959). 

90. Kornberg, A., Zimmerman, S. B., Kornberg, S. R., and Josse, J. Enzymatic synthesis 
of deoxyribonucleic acid. VI. Influence of bacteriophage T2 on the synthetic 
pathway in host cells. Proc. Natl. Acad. Sci. U.S., 43, 772 (1959). 

91. Krueger, A. P. The sorption of bacteriophage by living and dead susceptible 
bacteria. J. Gen. Physiol. 14, 493 (1931). 

92. Krueger, A. P., and Northrop, J. H. The kinetics of the bacterium-bacteriophage 
reaction. /. Gen. Physiol., 14, 223 (1931). 



Bibliography xxvii 

93. Krueger, A. P., and Scribner, E. J. Nature of intracellular phage precursor. /. Gen. 
Physiol., 22, 699 (1939). 

94. Latarjet, R. Multiplication of bacterial viruses studies by radiobiological methods, 
in "The Nature of Virus Multiplication," P. Fildes and W. E. van Heyningen, eds. 
(2nd Symp. Soc. Gen. Microbiol), p. 175. Cambridge University Press, Cambridge 
(1953). 

95. Lea, D. E. "Actions of Radiations on Living Cells." Cambridge University Press, 
Cambridge, (1947). 

96. Lederberg, E. M., and Lederberg, J. Genetic studies of lysogenicity in Escherichia 
coli. Genetics, 38, 51 (1953). 

97. Lederberg, J. Recombination mechanisms in bacteria. J. Cell. Conip. Physiol., 45, 
Supp. 2, 75 (1955). 

98. Lennox, E. S. Transduction of linked genetic characters of the host by bacteriophage 
PI. Virology, 1, 190 (1955). 

99. Lennox, E. S., Luria, S. E., and Benzer, S. On the mechanism of photoreactivation 
of ultraviolet inactivated bacteriophage. Biochim. et Biophys. Acta, 15, 471 (1954). 

100. Levinthal, C. Recombination in phage T2; its relation to heterozygosis and growth. 
Genetics, 39, 169 (1954). 

101. Levinthal, C. The mechanism of DNA replication and genetic recombination in 
phage. Proc. Natl. Acad. Sci. U.S., 42, 394 (1956). 

102. Levinthal, C. Bacteriophage genetics, in "The Viruses," F. M. Burnet and W. M. 
Stanley, eds. Vol IL, p. 281. Academic Press, New York, 1959. 

103. Levinthal, C, and Fisher, H. The structural development of a bacterial virus. 
Biochim. et Biophys. Acta, 9, 419 (1952). 

104. Litman, R. M., and Pardee, A. B. Production of bacteriophage mutants by a 
disturbance of dexoyribosenucleic acid metabolism. Nature, 178, 529 (1956). 

105. Luria, S. E. Mutations of bacterial viruses affecting their host range. Genetics, 30, 
84 (1945). 

106. Luria, S. E. Reactivation of irradiated bacteriophage by transfer of self-reproducing 
units. Proc. Natl. Acad. Sci. U.S., 33, 253 (1947). 

107. Luria, S. E. Bacteriophage: An essay on virus reproduction. Science, 111, 507 
(1950). 

108. Luria, S. E. The frequency distribution of spontaneous bacteriophage mutants as 
evidence for the exponential rate of phage reproduction. Cold Spring Harbor Symp. 
Quant. Biol, M, 463 (1951). 

109. Luria, S. E. Host-induced modifications of bacterial viruses. Cold Spring Harbor 
Symp. Quant. Biol., 18, 237 (1953). 

110. Luria, S. E. Radiation and viruses, in "Radiation Biology," A. HoUaender, ed. 
Vol. 11, p. 333. McGraw Hill Book Co., New York, 1955. 

111. Luria, S. E., and Delbruck, M. Interference between bacterial viruses: IL Inter- 
ference between inactivated bacterial virus and active virus of the same strain and 
of a different strain. Arch. Biochm., 1, 207 (1942). 

112. Luria, S. E., and Dulbecco, R. Genetic recombinations leading to production of 
active bacteriophage from ultraviolet inactivated bacteriophage particles. Genetics, 
34, 93 (1949). 



xxviii Introduction 

1 13. Luria, S. E., and Latarjet, R. Tltraviolet irradiation during intracellular growth. 
J. BaderioL, 53, 149 (1947). 

114. Lwoff. A. Lysogeny. BaderioL Rev., 17,2m {\<db'i). 

115. Lwoff, A. Bacteriophage as a model of host-virus relationship, in "The Viruses," 
F. M. Burnet and W. M. Stanley, eds. Vol. II, p. 187. Academic Press, New York, 
1959. 

116. Lwoff, A., and Gutmann, A. Recherches sur un Bacillus megatherium lysogene. 
Ann. Inst. Pasteur, 78,111 (1950). 

117. Lwoff, A., Siminovitch, L., and Kjeldgaard, N. Induction de la lyse bact^rio- 
phagique de la totality d'une population microbienne lysogene. Compt. rend. 231, 
190 (1950). 

118. Lwoff, A., Siminovitch, L., and Kjeldgaard, N. Induction de la production de 
bacteriophages chez une bacterie lysogene. Ann. Inst. Pasteur, 79, 815 (1950). 

119. Maal0e, 0., and Symonds, N. Radioactive sulfur tracer studies on the reproduction 
of T4 bacteriophage. J. BaderioL, 65, 177 (1953). 

120. Maal0e, 0., and Watson, J. D. The transfer of radioactive phosphorus from parental 
to progeny phage. Proc. NatL Acad. Sci. U.S., 37, 507 (1951). 

121. McKinley, E. B. S^rum antilytique obtenu par immunisation contre une bacterie 
normale. CompL rend. Sac. BioL, 93, 1050 (1925). 

122. Meselson, M., and Stahl, F. W. The replication of DNA in Escherichia coli. Proc. 
NatL Acad. Sci. U.S., U, 671 (1958). 

123. Meuli, H. Studien zum Bakteriophagen-problem. II. Die Konzentration des 
lytischen Prinzips und ihre Beziehung zum Ablauf der Bakteriophagenreaktion, 
Z. Hyg. Infektionskrankh., 99, 46 (1923). 

124. Northrop, J. H. Increase in bacteriophage and gelatinase concentration in cultures 
of Bacillus megatherium. J. Gen. Physiol. 23, 59 (1939). 

125. Puck, T. T. The first steps of virus invasion. Cold Spring Harbor Symp. Quant. 
BioL, 18, 149 (1953). 

126. Putnam, F. W. and Kozloff, L. M. Biochemical studies of virus reproduction. IV. 
The fate of the infecting virus particle. J. BioL Chem., 182, 243 (1950). 

127. Raettig, H. "Bakteriophagie, 1917 bis 1956." Gustav Fischer, Stuttgart, 1958. 

128. Rubin, H., and Temin, H. M. A radiological study of cell-virus interaction in the 
Rous sarcoma. Virology, 7, 75, (1959). 

129. Schlesinger, M. Ueber die Bindung des Bakteriophagen an homologe Bakterien. 

I. Die Unterscheidung von Gruppen von verschiedener Bindungsaffinitaet innerhalb 
der Bakterien des selben Lysats. Die Frage der Reversibilitaet oder Irreversibilitaet 
der Bindung. Z. Hyg. Infektionskrankh., 114, 136 (1932). 

130. Schlesinger, M. Ueber die Bindung des Bakteriophagen an homologe Bakterien. 

II. Quantitative Untersuchungen ueber die Bindungsgeschwindigkeit und die 
Saettigung. Berechnung der Teilchengroesse des Bakteriophagen aus deren Ergeb- 
nissen. Z. Hyg. Infektionskrankh., 114, 149 (1932). 

131. Schlesinger, M. Reindarstellung eines Bakteriophagen in mit freiem Auge sichtbaren 
Mengen. Biochem. Z., 264, 6 (1933). 

132. Schlesinger, M. Zur Frage der chemischen Zusammensetzung des Bakteriophagen. 
Biochem. Z., 273, 306 (1934). 

133. Sertic, V. Proc^de d'obtention de variantes de bacteriophage adapt^es a lyser des 
formes bact^riennes secondaires. Compt. rend. Soc. BioL, 100, 612 (1929). 



Bibliography xxix 

134. Siegel, A., Ginoza, W., and Wildman, S. G. The early events of infection with 
tobacco mosaic virus nucleic acid. Virology, 3, 554 (1957). 

135. Siegel, A., and Wildman, S. G. The inactivation of the infectious centers of tobacco 
mosaic virus by ultraviolet light. Virology, 2, 69 (1956). 

136. Stahl, F. W. Radiobiology of bacteriophage, in "The Viruses," F. M. Burnet and 
W. M. Stanley, eds., Vol. II, p. 353. Academic Press, New York, 1959. 

137. Steinberg, C., and Stahl, F. The theory of formal phage genetics. Cold Spring 
Harbor Symp. Quant. Biol, 23, 42 (1958). 

138. Stent, G. S. Decay of incorporated radioactive phosphorus during reproduction of 
bacteriophage T2. /. Gen. Physiol., 38, 853 (1955). 

139. Stent, G. S. Mating in the reproduction of bacterial viruses. Advances in Virus 
Research, 5, 95 (1958). 

140. Stent, G. S. Intracellular multiplication of bacterial viruses, in "The Viruses," 
F. M. Burnet and W. M. Stanley, eds.. Vol. II, p. 237. Academic Press, New York, 
1959. 

141. Stent, G. S., and Fuerst, C. R. Inactivation of bacteriophages by decay of incor- 
porated radioactive phosphorus. J. Gen. Physiol., 38, 441 (1955). 

142. Stent, G. S., and Fuerst, C. R. Genetic and physiological effects of the decay of 
incorporated radioactive phosphorus in bacterial viruses and bacteria. Advances in 
Biol, and Med. Physics, 7, 1 (1960). 

143. Stent, G. S., and Jerne, N. K. The distribution of parental phosphorus atoms among 
bacteriophage progeny. Proc. Natl. Acad. Sci. U.S., 41, 704 (1955). 

144. Stent, G. S., and Maal0e, 0. Radioactive phosphorus tracer studies on the reproduc- 
tion of T4 bacteriophage. II. Kinetics of phosphorus assimilation. Biochim. el 
Biophys. Acta., 10, 55 (1953). 

145. Stent, G. S., Sato, G. H., and Jerne, N. K., Dispersal of the parental nucleic acid of 
bacteriophage T4 among its progeny. J. Mol. Biol., /, 134 (1959). 

146. Stent, G. S., and WoUman, E. L., On the two-step nature of bacteriophage adsorp- 
tion. Biochim. et Biophys. Acta, 8, 260 (1952). 

147. Stocker, B. A. D., Zinder, N. D., and Lederberg, J. Transduction of flagellar 
characters in Salmonella, J. Gen. Microbiol., 9, 410 (1953). 

148. Tolmach, L. J. Attachment and penetration of cells by viruses. Advances in Virus 
Research, 4, 63 (1957). 

149. Twort, F. W. An investigation on the nature of the ultramicroscopic viruses. 
Lancet, 1915, II, 1241. 

150. Visconti, N., and Delbruck, M. The mechanism of genetic recombination in phage. 
Genetics, 38, 5 (1953). 

151. Watson, J. D., and Crick, F. H. C. The structure of DNA. Cold Spring Harbor 
Symp. Quant. Biol, 18, 123 (1953). 

152. Watson, J. D., and Maal0e, O. Nucleic acid transfer from parental to progeny 
bacteriophage. Biochim. et Biophys. Acta, 10, 432 (1953). 

153. Weidel, W. "Virus," Springer Verlag, Berlin, 1957. (University of Michigan Press, 
Ann Arbor, 1959.) 

154. Weidel, W. Bacterial viruses (with particular reference to adsorption/ penetration). 
Ann. Rev. Microbiol. 12, 27 (1958). 

155. Williams, R. C., and Fraser, D. Structural and functional differentiation in T2 
bacteriophage. Virology, 2, 289 (1956). 



XXX 



Introduction 



156. Wollman, E. Bact^riophagie et processus similaires. H6r6dit6 ou infection? Bull. 
Inst. Pasteur, 26, 1 (1928). 

157. Wollman, E., and Wollman, E. R6g6n^ration des bacteriophages chez le B. mega- 
therium lysogene. Compt. rend. Soc. Biol, 122, 190 (1936). 

158. Wollman, E. L. Sur le determinisme g^n^tique de la lysogenie. Ann. Inst. Pasteur, 
84, 281 (1953). 

159. Wollman, E. L., and Jacob, F. Lysogenie et r^combinaison g^n^tique chez Escheri- 
' chia coli K12. Compt. rend., 239, 455 (1954). 

160. Wollman, E. L., and Jacob, F. Sur les processus de conjugation et de r^combinasion 
chez Escherichia coli. II. La localisation chromosomoqie du prophage et les conse- 
quences de I'induction zygotique. Ann. Inst. Pasteur, 93, 323 (1957). 

161. Wyatt, G. R., and Cohen, S. S. The bases of the nucleic acids of some bacterial and 
animal viruses: The occurrence of 5-hydroxymethylcytosine. Biochem. J., 65, 774 
(1953). 

162. Wyckoff, R. W. G. The electron microscopy of developing bacteriophage. II. 
Growth of T4 in liquid culture. Biochim. et Biophys. Acta., 2, 246 (1948). 

163. Zinder, N. D. Infective heredity in bacteria. Cold Spring Harbor Symp. Quant. 
Biol., 18, 261 (1953). 

164. Zinder, N. E., and Lederberg, J. Genetic exchange in Salmonella. J. Bacteriol., 64, 
679 (1952). 



PAPERS ON BACTERIAL VIRUSES 



NINETIETH ANNUAL MEETING 

OF THE 

Held at Glasgow, July, 1922. 



DISCUSSION ON 
THE BACTERIOPHAGE (BACTERIOLYSIN). 



I.-THE NATURE OF BACTERIOPHAGE. 

By Dr. F. d'Herelle. 

( From the Pasteur Institute, Paris. ) 

Let us record first, in a few words, the essential facts relating to the present 
discussion. 

There exist in the intestinal contents of all living beings principles which 
have the property of dissolving certain bacteria. These principles pass 
porcelain filters; hence it is possible to separate them from intestinal bacteria. 

Each filtering principle, isolated from intestinal contents of normal indi- 
viduals, dissolves a certain number of bacterial species, belonging generally to 
the coli-typhoid-dysentery group. The action, therefore, of these principles is 
not strictly specific; but belongs, for a given filtrate, to a certain number of 
microbial species. In convalescents from an infectious disease a principle 
endowed with an energetic dissolving action on the bacterial species which 
causes the disease is always met with. 

It is not only towards the intestinal bacteria that these lytic principles 
exist: for instance, one can isolate from the excreta of convalescents from 
bubonic plague a principle which dissolves B. pestis; the same phenomenon 
exists in various animal septicaemias. 

I cannot enter here into an examination of all the facts relating to bacterio- 
phage,^ facts most of which have been already confirmed by the investigations 
of other workers. But if the facts themselves have not been contested, it is not 
so in regard to the hypothesis which I have put forward to explain the nature 
of the principle which brings about the dissolution of bacteria. 

Reprinted by permission of the British Medical Association from 
the British Medical Journal, 2 (3216), 289-297, August 19, 1922. 



Aug. 19, 1922 ] SECTION OF MICROBIOLOGY. [MelrcAf Jo3al 

A preliminary remark presents itself. Many authors who have formulated 
hypotheses as to the nature of the lytic principle have simply taken a particular 
fact supporting their own point of view and have neglected the whole assembly 
of experimental facts which renders such a thesis untenable, thereby forgetting 
that experiment is the final criterion of the truth of a theory. Let us try to 
resolve the question objectively. 

The choice between the different hypotheses a priori possible to explain 
the origin of the lytic principle can only be made in accordance with a funda- 
mental rule of logic — namely, in order that a hypothesis becomes admissible 
it is necessary that it should explain all the experimental facts and that it is 
not contradicted by any of them; further, in order that this admissible 
hypothesis may be considered as being in conformity with the nature of things, 
it must be proved that all the experimental facts cannot be explained if one 
abandons or modifies this hypothesis. 

The bacteriolysis produced under the influence of the principle which we 
have named "bacteriophage" consists in a total dissolution of the microbial 
body; at the end of this action there remains no visible residue. A total dissolu- 
tion of a microbial substance can only be due to a transformation or decompo- 
sition of the proteins of the microbe by proteolytic enzymes. Whence, it may 
be asked, come these proteolytic enzymes? Four hypotheses may be con- 
sidered. 

1. The enzymes may be derived from the animal organism which is 
attacked by the given bacteria. The enzymes would then be the result of a 
defensive reaction on the part of the organism. 

This is the hypothesis of Kabeshima, Bordet and Ciuca, and Ann Kuttner. 
Kabeshima- does not specify the particular tissue of the animal from which the 
enzymes originate; Bordet and Ciuca'^ indicate the leucocytes; Ann Kuttner^ 
incriminates any tissue. 

2. The enzymes may come from intestinal bacteria as the result of a micro- 
bial antagonism. For instance, one knows the bacteriolytic action of filtrates 
of old cultures of B. pyocyaneiis. 

This is the hypothesis of Lisbonne et Carrere, ' for whom the lytic enzymes 
are secreted by intestinal bacilli such as B. coli, B. proteus, etc. 

3. The enzymes may be secreted by the bacterium itself which undergoes 
the lysis. These enzymes would therefore be autolysins. 

This is the hypothesis of Weinberg and Aznar.*' Under the action of a 
cause X the bacteria would acquire the property of secreting autolytic enzymes. 

BaiP had already indicated a similar origin, but his hypothesis was more 
complicated, and that in order to harmonize his conception with certain 
experimental facts to which we shall refer later. The bacteria undergoing lysis 
would become fragmented into filtrable corpuscles (the "splitters" hypothesis 
of Bail). These corpuscles would secrete enzymes capable of dissolving the 
bacteria from which they spring. 

With the same view, Otto and Winkler^ have put forward a hypothesis 



Aug. 19, 19221 THE BACTERIOPHAGE. [ Mrllclf jo":".:. 

closely related to that of Bail, but for them the question resolved itself into 
one of bacterial albuminoid micelles, which do not reproduce themselves, but 
which secrete the lytic enzymes. 

4. The enzymes may be secreted by an ultramicroscopic virus, which is a 
parasite of bacteria. This is the hypothesis by which I have held since my first 
publication." 

Those four hypotheses cover all the possibilities. Let us now see which 
hypothesis conforms best with the experimental facts, at the same time 
eliminating those which fail to do so. 

FmsT Fact: The dissolution of bacteria tinder the influence of the bacterio- 
phagic principle takes place in series. 

That is to say, a filtrate containing the bacteriophagic principle, when 
added to a culture in liquid medium of a bacterium towards which the 
principle manifests its lytic action, provokes the entire dissolution of the 
bacteria. A trace of the latter, inoculated into a second bacterial culture, leads 
to the latter in its turn undergoing total lysis. A trace of this second dissolved 
culture, inoculated into a third bacterial culture, produces the same phenom- 
enon, and so on ad infinitum. After more than a thousand passages, the 
thousandth bacteriolysed culture contains a bacteriophagic principle as active 
as, and generally much more active than, that of the primitive filtrate. 

The phenomenon goes on in series in the same way, whether the culture is 
inoculated with the previously dissolved non-filtered culture, or with the 
dissolved culture filtered through a porcelain filter: the filtrate contains the 
bacteriophagic principle as active as the lysed non-filtered culture. 

A simple enzyme action would cease to show itself from the first tubes of 
the series, because of the greater and greater dilution of the enzyme solution 
in the course of the successive passages. One calculates easily that at the 
thousandth passage (each passage being carried out on 10 c.cm. of bacterial 
emulsion inoculated with 1/1000 c.cm. of the preceding dissolved culture) the 
titre of the dilution in the thousandth tube of the series is given in cubic kilo- 
metres by the the number 10^"^-. To appreciate this incommensurable number 
it is interesting to note that, at the twenty-second passage only, the drop of 
primitive filtrate introduced into the first tube of the series finds itself diluted 
in 10^" cubic kilometres of liquid; such a cube of liquid is so great that a light 
ray would take one billion years to cross one edge of it. It is impossible to 
conceive of an enzyme, contained in a single drop of primitive filtrate, still 
existing, without any diminution of its activity, after being diluted to this 
extent. 

That the action may become manifest, each dilution must be allowed 
sufficient time (four to six hours) to permit the development of the bacterio- 
phagic principle. On the contrary, when the dilution is utilized immediately, 
the bacteriophagic action ceases about the fourth or fifth dilution. 

These experiments establish clearly that a regeneration of the bacterio- 
phagic principle occurs at each passage. Consequently the lytic enzymes are 



Aug. 19, 1922 ] SECTION OF MICROBIOLOGY. [ MnlZ^YoT"... 

produced by a living being, which regenerates itself in the course of the 
successive passages. 

It will be obvious, then, that this transmissibility in series eliminates 
hypotheses 1 and 2. The authors who have put forward these hypotheses must 
admit that the principle derived from the organism which is attacked by para- 
sites (hypothesis 1), or from the foreign bacteria (hypothesis 2), plays solely 
the role of a transformer— that is to say, it gives rise in the bacteria which 
undergo the transmissible lysis to a special state of "autolysability." We 
come now to the third hypothesis. Only Bordet and Ciuca have attempted to 
give an explanation of this. For them the leucocytic principle would provoke 
in the bacteria a "hereditary vitiation of nutritional nature." I must say that 
I do not understand how a filtered liquid can transmit a hereditary property. 
Besides, the whole theory of Bordet and Ciuca is based on an experience 
entitled "leucocytic exudates," which I have elsewhere stated it is impossible 
to repeat— a statement which has not been challenged by Bordet." The result 
obtained by him in one experiment would appear to be purely accidental." 

But it is to no purpose to discuss here, point by point, the di£Ferent theories; 
it is sufficient to show that they are contrary to the facts. Hypotheses 1 and 2 
being eliminated, there only remain to consider hypotheses 3 and 4. 

Second Fact: The lytic enzymes emanate from material corpuscles which 
traverse filters; these corpuscles multiply in the course of the bacteriolysis. 

Experiment A. 

hi a well-developed bacterial culture, showing a marked turbidity, let us add 
a small quantity (say 1/10,000,000 c.cm.) of a filtrate containing the bacteriophagic 
principle, shake, withdraw 1 20 c.cm. and spread it over the surface of an agar slope. 
Thus we spread over the agar a great number of bacteria and a very small quantity of 
bacteriophagic principle. After incubation (eighteen to t\venty-four hours at 37°) 
we note that the surface of the agar shows a bacterial growth with a certain number 
of circular bare spaces here and there, where the agar presents no trace of growth. 
These vacant spaces once formed are unchangeable; they never spread, and they 
are never invaded by the surrounding culture. 

It is the presence of these immutable bare spaces, which are perfectly 
circular, that characterizes what we have named "bacteriophage." Before 
concluding that any bacteriolytic phenomenon is bacteriophagic, it becomes 
necessary to verify if it gives rise to such bare spaces on agar culture. If not, 
the bacteriolysis is not of bacteriophagic nature. 

The number of spaces depends simply on the quantity of filtrate added to 
the bacterial culture. If into various bacterial emulsions we introduce variable 
quantities of the filtrate, the number of bare spaces is strictly proportional to 
the quantity of filtrate added. On the other hand, the number of the bare 
spaces is independent of the number of bacteria contained in the medium; 
whether one introduces a given quantity of filtrate into a bacterial emulsion 
containing one hundred millions or ten billions of bacteria per cubic centi- 
metre, the number of bare spaces on the agar is practically always the same. 

The phenomenon of these vacant spaces is only comprehensible on the 



Aug. 19, 1922] THE BACTERIOPHAGE. [ M Jrcif Jor;.. 

supposition that the bacteriophagic principle is of corpuscular nature. The 
following experiments give the proof thereof. 

Experiment B. 

Given a filtrate containing the bacteriophagic principle, let us determine its 
force in the production of bare spaces. 

It is well known that to determine the number of living bacteria contained in 
an emulsion agar plates are inoculated with a dilution of this emulsion; after incuba- 
tion, the number of colonies which develop multiplied by the titre of the dilution 
gives the number of bacteria per cubic centimetre of the primitive emulsion. The 
enumeration of the bacteriophagic corpuscles contained in a filtrate is estimated in 
exactly the same way. But as the bacteriophage only grows at the expense of living 
bacteria, we must make the dilutions of the filtrate in a bacterial emulsion. For 
this we pipette over agar the bacterial emulsion containing a given quantity of 
filtrate: after incubation we obtain a bacterial layer strewn with circular bare spaces, 
each of these spaces being a colony of bacteriophage issued from one corpuscle. 
The number of bare spaces multiplied by the titre of the dilution gives the number 
of ultramicroscopic bacteriophagic corpuscles contained in 1 c.cm. of the primitive 
filtrate. 

This experiment shows that the behaviour of the bacteriophage is exactly 
the same as that of any ordinary microbe. But this last develops at the expense 
of the nutritive substances contained in the medium; the bacteriophage 
develops at the expense of the bacterial bodies which constitute its nutritive 
medium. The bare spaces represent places cleared up by the growth of the 
ultramicroscopic bacteriophagic corpuscles. 

Experiment C. 

Now dilute a filtrate so as to obtain a dilution such that 1 c.cm. contains one 
bacteriophagic corpuscle. Dilute this 1 c.cm. with 9 c.cm. of sterile water, and 
inoculate ten tubes of bacterial emulsion each with 1 c.cm. It will be obvious that 
only one of the ten tubes will contain the generator of a bare space; the nine others 
will not contain any. Place the ten tubes in the incubator at 37° for twenty-four 
to forty-eight hours; it will be seen that only one of the ten microbial emulsions 
shows bacteriob'sis; the nine others will remain unchanged, the bacteria remain liv- 
ing, normal, and subculturable. 

The lytic action, therefore, is complete when only one generator of a bare 
space is introduced into a bacterial emulsion; the action is nil in the contrary 
case. This experiment can only be explained on the supposition that the 
bacteriophagic principle, the source of the lytic enzymes, is a corpuscle; and 
that each corpuscle deposited on the agar in the midst of the bacteria gives 
rise to a colony of these ultramicroscopic corpuscles, such a colony being repre- 
sented by a bare space. 

Experiment D. 
Inoculate a bacterial emulsion with a bacteriophagic filtrate of known corpuscular 
strength (that is, of known bare spaces forming force); make agar slopes every hour 
in the same way as in Experiment A. In this way, after six hours, we will have six 
agar slopes inoculated, and, on incubation, it will be seen that the number of 
bacteriophagic spaces (each bare space corresponding to a corpuscle) increases in 
proportion as the lytic action progresses in the emulsion. The maximum number of 
bare spaces is given when the lysis of the original emulsion is complete; at this 



Aug. 19, 1922 ] SECTION OF MICROBIOLOGY. [ ulZ^Zl.. 

moment there are no longer any living bacteria in the emulsion— it is a pure culture 
of the ultramicroscopic bacteriophagic corpuscles. The bacteriophagic power of such 
a lysed culture remains at its maximum level during many weeks— it is practically 
"fixed." 

The foregoing experiments show that the lytic enzymes are originated by 
ultramicroscopic corpuscles which multiply and reproduce themselves. Hence 
those hypotheses explaining the phenomenon as being due to a soluble auto- 
lytic enzyme become eliminated. 

Of all the hypotheses put forward, only that of Bail, and perhaps that of 
Otto and Winkler, could not be incompatible with these experiments, since 
they suppose that the element which secretes the lytic enzymes can be 
furnished by ultra-microscopic corpuscles, but derived from the bacteria. It 
is difficult to reconcile these hypotheses with the fact that the bacteriophagic 
action is not specific, and that the same strain of bacteriophage can dissolve 
bacteria of different species. Indeed, if one can understand how a principle 
originated by a species of bacteria can provoke the formation of a similar 
principle in a culture of the same bacterial species, it is hard to admit how this 
principle could provoke the formation of the bacteriophagic principle in a 
culture of a different bacterial species. Besides, these two hypotheses are 
rendered untenable by the following fact. 

THrao Fact: All bacteriophagic ultramicroscopic corpuscles, grown at the 
expense of any bacterial species, constitute one and the same antigen. 

I have up to the present isolated various strains of bacteriophage active 
towards different bacteria: B. typhosus, B. paratyphosus A and B, B. dysen- 
teriae (Shiga, Flexner, Hiss), B.coli, B.pestis, B.proteus, B. gallinarum, bac- 
teria of haemon-hagic septicaemias, staphylococcus, etc. 

We have seen above what constitutes a culture of the ultramicroscopic 
bacteriophagic corpuscles. Inoculate an emulsion of B. dysenteriae with a 
filtrate containing an active bacteriophage towards this bacterium; after a few 
hours the medium becomes limpid, the bacilli are dissolved, and the bacterio- 
phage corpuscles have multiplied; the culture of the bacteria has become a 
culture of antidysenteric bacteriophage. Similarly, a culture of B. pestis 
dissolved under the action of the bacteriophagic principle (isolated from the 
intestinal contents of a convalescent from bubonic plague) becomes a culture 
of antipestic bacteriophage. And so on for any other microbe. 

One knows that the serum of a rabbit prepared by injections of B. dysen- 
teriae contains an amboceptor towards B. dysenteriae, but does not contain any 
amboceptor towards any other bacteria— for example, B. pestis. 

Now prepare a rabbit with a culture of B. dysenteriae dissolved by bac- 
teriophagic action; one verifies that its serum does not contain any amboceptor 
towards normal B. pestis, but contains an amboceptor for B. pestis dissolved 
by the action of the bacteriophage. And this is true for no matter what bac- 
terial species; the serum of an animal prepared by a culture of any bacterial 
species dissolved by bacteriophage contains an amboceptor which fixes itself 
on any other bacteriophaged culture. Hence the amboceptor is specifically 



Aug. 19, 1922] THE BACTERIOPHAGE. [ M Jlcl^ Jo-kL. 

antibacteriophagic, and not antibacterial. Therefore the bacteriophage is an 
autonomous antigen, and consequently must be considered, in the present 
state of knowledge, as a definite ultramicroscopic virus constituted by cor- 
puscles, such a virus being necessarily an ultramicrobe. 

Only the hypothesis of a liltrable micro-organism which is an obligatory 
parasite of bacteria is compatible with the foregoing facts. It now remains to 
show that this only admissible hypothesis is not contradicted by any experi- 
mental fact. 

Fourth Fact: The ultramicroscopic corpuscles possess a variable virulence. 

We have seen that the above ultramicroscopic organism, which we have 
designated by the name of "bacteriophage," is constituted by living corpuscles 
which multiply and which it is possible to count. Now, experience shows that 
the lytic activity of these corpuscles varies from one strain to another. For 
instance, certain strains recently isolated from the organism are unable to 
provoke the dissolution of a bacterial emulsion, and no matter what number 
of corpuscles are inoculated into this bacterial emulsion, one can only perceive 
the presence of these bacteriophagic corpuscles by the production of pin-point 
bare spaces on agar. On the contrary, with very active strains giving on agar 
large bare spaces (4 to 5 mm. in diameter), the inoculation of one corpuscle 
is sufficient to provoke a total dissolution of all the bacteria contained in the 
emulsion. It is easy to prove that this difference in action is due to a difference 
in the multiplication of the corpuscles inoculated. In the case of feeble strains 
reproduction is slow; in the case of strong strains it is rapid. The active 
particle, therefore, is endowed with a variable virulence, the term "virulence" 
being taken in its real bacteriological sense of "vegetative power in vivo." To 
explain this fact one must consider the active corpuscle as a microbe, because 
only a microbe possesses a variable virulence from strain to strain. Moreover, 
an enzyme only acts by its quantity; a microbe by its virulence and its toxicity 
combined. 

Fifth Fact: By successive passages, it is possible to increase the virulence 
of a feeble strain of bacteriophage. 

For instance, inoculate a feeble strain of bacteriophage into a culture of 
B. typhosus; filter the mixture after incubation; introduce a drop of this filtrate 
into a fresh culture of B. typhosus, which is filtered in turn, and so on for a few 
passages. The virulence is found to increase little by little, and after a certain 
number of passages the virulence becomes such that a trace of the filtrate is 
sufficient to provoke the total dissolution of a fresh emulsion of this bacillus. 
The technique of the exaltation of virulence is really the same as for any 
microbe: we exalt the virulence of a bacterium for a given animal by serial 
passages in this animal species; we exalt the virulence of the bacteriophage for 
a given bacterial species by serial passages in this bacterium. This fact is only 
compatible with the hypothesis that bacteriophage is a microbial parasite, 
since adaptation is the prerogative of living beings. 



Aug. 19, 1922 ] SECTION OF MICROBIOLOGY. [ M.IrJlf JoTn.. 

Sixth Fact: Experience shows that bacteria attacked by bacteriophage do 
not remain passive; they defend themselves and are even capable, under certain 
conditions, of acqtdring an immunity towards the parasite. 

This defence of the bacteria manifests itself by a double mechanism: the 
bacteria surround themselves with a capsule, and secrete "aggressines" capable 
of paralysing the lytic enzymes of the bacteriophage. Only the hypothesis of 
bacteriophage being a parasite of bacteria conforms with these facts. 

Seventh Fact: The behaviour of bacteriophage towards physical and 
chemical reagents is that of a living being, and does not agree with that of 
an enzyme. 

The resistance of bacteriophage to the action of physical and chemical 
reagents is intermediate between that of the vegetative and the spore forms of 
ordinary bacteria. The resistance is less than that of certain ultramicrobes, 
notably that of the tobacco mosaic. It is destroyed at 74°-75°C., but ceases 
to develop at 43°. Between 43° and 74° it possesses no lytic action. As to 
the action of the antiseptics, the bacteriophage is killed after twenty-four 
hours' contact with a 1 per cent, solution of a neutral salt of quinine, after 
forty-eight hours' contact with 95 per cent, alcohol, and by eight days' contact 
with glycerin. The last-named liquid is precisely the medium employed for 
preserving indefinitely an enzyme in soluble form. 

Eighth Fact: It is possible to extract the lytic enzymes free from the living 
bacteriophage mirco-organism. 

If to one part of a culture of bacteriophage is added nine parts of absolute 
alcohol and the mixture left forty-eight hours in contact, a precipitate is 
obtained which contains bacteriolytic enzymes, while the bacteriophagic germs 
are killed, which is proved by the failure of transmissibility in series of the 
phenomenon. One can only understand this fact on the supposition of an 
ultramicroscopic microbial parasite: the bacteria are dissolved by enzymes 
secreted by the living ultramicroscopic germs, hence the action in series. In 
the above experiment the ultramicroscopic germs are killed, while the enzymes 
already secreted are preserved intact; hence only a lytic action is obtained 
which is not transmissible in series. 

Ninth Fact: Bacteriophage is capable of adaptation. 

The bacteriophage is very sensitive to the action of acids— much more 
sensitive than ordinary microbes. It can be accustomed progressively to live 
in media containing quantities of an acid which would have been destructive 
beforehand. The same fact is observed with glycerin. Adaptation is a preroga- 
tive of living things. 

Tenth Fact: The properties of bacteriophage are essentially variable. 

It is impossible to isolate two strains of bacteriophage which are absolutely 
identical as to their range of action on diflFerent bacterial species, and as to 
their intensity of action towards each of these bacterial species. For a given 



10 



Aug. 19, 1922] THE BACTERIOPHAGE. [ MrlrclXT; .. 

strain of bacteriophage one can vary experimentally this intensity. Variability 
is an essential characteristic of life. 

We have seen, first, that of all the possible hypotheses concerning the 
nature of the bacteriophage, the hypothesis of an ultramicroscopic parasite of 
bacteria is alone admissible. Moreover, we have just seen that this hypothesis 
is not contradicted by any of the facts of bacteriophagic phenomena; and, 
furthermore, all of the facts cannot be explained if this hypothesis is aban- 
doned or modified. It follows logically, therefore, that such a hypothesis be- 
comes a certitude. 

Further, I would remark that I do not specify in any way the species to 
which the ultramicroscopic organism, which I have given the name of Bac- 
teriophogtwi intestinole, belongs; the name simply recalls its characteristic 
property and the place where it was at first found. Is it a protozoon, a fungus, 
a bacterium? Does it belong to a kingdom which is neither the vegetable nor 
the animal— a still simpler form of life than any which we at present know? 
These are questions which cannot at present be answered. All that we know 
of it is that it is an ultramicroscopic organism, a filtrable being, parasite of 
bacteria, endowed with functions of assimilation and reproduction— functions 
which characterize the living nature of the being which possesses these proper- 
ties. That is all that experiment actually shows us.* 

The foregoing discussion can only apply to the phenomenon of bacteriolysis 
in series presenting the special characters that I have indicated. This being so, 
a final question arises— that of the plurality of the serial phenomena. 

In 1915 Twort^- described a bacterial transformation taking place in series, 
under the influence of a principle which passes through porcelain filters. His 
researches have been made chiefly with a micrococcus isolated from vaccine. 
In addition he observed an identical phenomenon with two other bacteria— a 
large indeterminate bacillus, and a bacillus of the coli-typhoid group. Twort 
isolated the active principle in the following manner. 

"Some interesting results, however, were obtained with cultivations from 
glycerinated calf vaccinia. Inoculated agar tubes, after twenty-four hours at 37°, 
often showed watery-looking areas, and in cultures that grow micrococci it was 
found that some of these colonies could not be subcultured, but if kept became 
glassy and transparent. On examination of these glassy areas nothing but minute 
granules, staining reddish with Giemsa, could be seen." 

The transforming principle is contained in this transparent material. 

"The transparent material when diluted (one in a million) with water or saline 
was found to pass the finest porcelain filters with ease, and one drop of the filtrate, 
pipetted over an agar tube, was sufficient to make that tube unsuitable for the growth 
of the micrococcus. That is, if the micrococcus was inoculated down the tube as a 
streak, this would start to grow, but would soon become dotted with transparent 
points which would rapidly extend over the whole growth. If in an infected tube 

* I will mention, in passing, the hypothesis of Salimbeni: the enzymes would be secreted 
by a myxobacteria, microscopically and even macroscopieally visible, given filtrable spores. 
All the workers who have studied the question have failed to observe this would-be 
myxobacteria, which besides, according to the description of Salimbeni, presents a mycelium 
(?). That can only mean an impurity, as I have elsewhere pointed out. 



11 



Aug. 19, 1922 ] SECTION OF MICROBIOLOGY. [ MBlrclX'L. 

small areas of micrococci are left— and this usually happens when the micrococcus 
has grown well before being infected— these areas will start to grow again and 
extend ov^er the transparent portion. . . ." 

According to this description of Twort, it is not a question of a real bac- 
terial dissolution, but a transformation of a normal culture on agar into a 
glassy and transparent one. This phenomenon is totally diflFerent from that 
produced by Bacteriophagum intestinale. Indeed, under the influence of this 
latter— no matter what bacterial species is dealt with, no matter what virulence, 
feeble or strong, of the strain of the bacteriophage, operating exactly in the 
same experimental conditions as Twort— in no circumstances does one observe 
the formation of transparent material or anything presenting such an appear- 
ance. 

Pipette over an agar tube a drop of a diluted filtrate containing a bacterio- 
phage active towards staphylococcus; then inoculate this tube as a streak with 
a pure culture of staphylococcus; after incubation we obtain a surface growth 
of the organism, which macroscopically, microscopically, and biologically is 
normal. This normal bacterial culture is strewn here and there with circular 
clear spaces; in the interior of these spaces the agar is bare, without any 
visible trace of any material. These bare spaces never spread over the sur- 
rounding culture, even after months in the incubator. Furthermore, they 
never become invaded by the surrounding bacterial culture: once formed, the 
bare spaces remain immutable. Confusion, therefore, between the phenomenon 
observed by Twort and the phenomenon provoked by the bacteriophage is 
in no way possible. 

What is the nature of the principle which acts in the phenomenon of bac- 
terial transformation observed by Twort? From a consideration of the observa- 
tions made on this phenomenon, and in view of the fact that the lytic agent 
and the bacteria are destroyed at the same temperature, it is probable that it 
is derived from the bacterium itself, which is capable of splitting up into 
fragments. Twort himself favours this view, but further experiments will be 
necessary to settle this question. 

This conclusion, however, does not detract in any way from the interest 
attaching to the important researches of Twort. One can already see that the 
phenomenon observed by this author may play an important role in the 
etiology of the so-called filtrable virus diseases, as he indeed seems to have 
foreseen. On the other hand, the bacteriophage undoubtedly plays a part in 
the defence of the organism in the course of infectious diseases, as I have 
shown elsewhere. 

From a survey of the results already arrived at in the domain of scientific 
investigation, it is easy to foresee that there exists a whole series of phenomena, 
quite unsuspected a few years ago, whose study should play a large part in 
the advance of scientific medicine. 

References. 

^For details see: Le Bacteriophage, son role dans I'immunite, in the collection, Mono- 
graphs of the Pasteur Institute ( Masson, Paris ) , of which an English edition is in prepara- 
tion (Williams and Wilkins, Baltimore, publishers). -Sur un ferment d'immunite bacteriol- 



12 



Aug. 19, 1922] THE BACTERIOPHAGE. [MBlrcAfjoTK.u 

ysant, C.R. Soc. Biologie, Ixxxiii, February 28th, 1920, p. 219. *^Le bacteriophage de 
d'Herelle, sa reproduction, son interpretation, C.R. Soc. Biologie, Lxxxiii, October 9th, 1920, 
p. 1296. 'On the influence of the tissue enzymes on the bacteriophage principle. Proceed. 
Exp. Biol, and Med., xviii, April 20th, 1 21, p. 222. sAntagonisme microbien et lyse trans- 
missible, C.R. Soc. Biologie, Ixxxvi, March 18th, 1922, p. 569 ''Autobacteriolysne et le 
phenomene de d'HereUe, C.R. Soc.Biologie, Ixxxvi, April 29th, 1922, p. 833. '^Bakerio- 
phagen Wirkungen gegen Flexner- imd Koli-Bakterien, Wien. klin. Woch., xxxiv. No. 37, 
September 15th, 1921. '^Uber die Natur des d'Herelle'schen Bakteriophagen, Deut. vied. 
Woch., No. 21, 1922. ^'Sur un microbe invisible, antagoniste des bacilles dysenteriques, 
C.R. Acad, des Sciences, clxv, September 10th, 1917, p. 373. ^o^xsudats leucocytaires et 
autolyse microbienne transmissible, C.R. Soc. Biologie, Ixxxiii, October 9th, 1920, p. 1293. 
"L'ultramicrobe bacteriophage, C.R. Soc. Biologie, October 29th, 1921, p. 721. i-An 
investigation on the nature of ultramicroscopic viruses. Lancet, December 4th, 1915. 



II.-THE BACTERIOPHAGE: THE BREAKING DOWN 
OF BACTERIA BY ASSOCIATED FILTER- 
PASSING LYSINS. 

By F. W. TwoRT, M.R.C.S., L R.C.P. 

( From the Laboratories of the Brown Institution, London. ) 

This phenomenon of bacteriolysis, which in France has been called the "bac- 
teriophage," consists, as you know, of a breaking down and dissolving up of 
bacteria by a filter-passing material which in certain circumstances may be 
associated with pure cultures. My first experiments were carried out at the 
Brown Institution during 1914 and 1915, for the Local Government Board, and 
the results were published in the Lancet (December 4th, 1915) under the 
title of "An investigation on the nature of ultramicroscopic viruses." The con- 
dition, in my opinion, is distinct from the various degenerative changes which 
have been so often described in bacterial cultures, and it will be necessary 
for me to consider in detail experiments on degenerative changes. It is also 
impossible in the time at my disposal to discuss every aspect of the subject; 
moreover, other speakers will no doubt deal with the many interesting experi- 
ments which they have carried out. In this paper I propose to give you as 
shortly as possible my original experiments as they were published in 1915, 
and I shall pass on to consider certain aspects of the subject which have since 
become points of controversy between different workers, and shall then deal 
shortly with certain more recent experiments which I have been doing on 
the biology of bacteria, as these may throw some light on the use of the lytic 
material. 

When starting the research my object was to discover, if possible, the na- 
ture and life-history of the ultramicroscopic group of viruses. The experiments 
were carried out with these views in my mind: In the first place we do not 
know for certain the nature of an ultramicroscopic virus. It may be a minute 
bacterium that will grow only on living material, or it may be a tiny amoeba 
which, like ordinary amoebae, thrives on living bacteria or animal tissue. On 
the other hand, it must be remembered that if the living organic world has 



13 



Aug. 19, 1922 ] SECTION OF MICROBIOLOGY. [MBlrc^XKNAL 

been built up slowly in accordance with the theories of evolution, then an 
amoeba and a bacterium must be recognized as highly developed organisms in 
comparison with much more primitive forms which once existed, and probably 
still exist, in nature. It is quite possible that an ultramicroscopic virus belongs 
somewhere in this vast field of life where the organization is lower than that 
of the bacterium or amoeba. It may be living protoplasm that forms no 
definite individuals, or an enzyme with power of growth. 

In the first instance attempts were made to demonstrate the presence of 
non-pathogenic filter-passing viruses. As is well known, in the case of ordinary 
bacteria, for every pathogenic micro-organism discovered many non-patho- 
genic varieties of the same type have been found in nature, and it seems highly 
probable that the same rule will be found to hold good in the case of ultra- 
microscopic viruses. It is difficult, however, to obtain proof of their existence, 
as pathogenicity is the only definite evidence we have at the present time of 
the presence of an ultramicroscopic virus. On the other hand, it seems 
probable that if non-pathogenic varieties exist in nature these should be more 
easily cultivated than the pathogenic varieties. 

The first experiments were carried out with such materials as soil, dung, 
grass, hay, straw, and with water from ponds. Cultivations were made on a 
large number of special media and under special conditions. Experiments 
were also carried out with pathological material obtained from distemper in 
dogs, from vaccinia, and from various other sources. The first results of 
interest were obtained with vaccinia. Inoculations were made on to ordinary 
agar tubes, and on to tubes of special egg media such as I used in my experi- 
ments for the cultivation of Johnes's bacillus and the lepra bacilli of man and 
of rats. It will be impossible to describe all these in detail, but the essential 
part of one series consists in the incorporation in the media of the dead bodies 
of certain acid-fast bacilli such as Bacillus phlei, which proved to be so success- 
ful in the cultivation of Johnes's bacillus. The egg media usually grew a 
number of colonies of micrococci and diphtheroids, while the agar tubes grew 
only a few colonies of micrococci. In the case of the egg media it was noted 
that in a few days certain parts of the micrococcus growth sometimes became 
dull, and in appearance rather resembled bacterial growth in which amoebae 
were also growing. On the agar media the colonies of micrococci ocassionally 
showed a translucent or transparent change, which started as more or less 
clear spots at the margins of the colonies. It was also found that some of these 
colonies could not be subcultured, but if kept the transparent change extended 
over most of the colony. 

On examination of these glassy areas nothing but minute granules, staining 
reddish with Giemsa, could be seen. Further experiments showed that if a 
colony of the white micrococcus that had started to become transparent was 
plated out instead of being subcultured as a streak, then the micrococcus grew 
and a pure streak culture from certain of these colonies could be obtained. On 
the other hand, if the plate cultures (made by inoculating the condensation 



14 



Aug. 19, 19221 THE BACTERIOPHAGE. [ MnlrcA^ Journal 

water of a series of tubes and floating this over the surface of the medium) 
were left, the colonies, especially in the first dilution, soon started to turn 
transparent, and the micrococci were replaced by fine granules. This action, 
unlike an ordinary degenerative process, started from the edge of the colonies; 
and further experiments showed that when a pure culture of the white or the 
yellow micrococcus isolated from vaccinia is touched with a small portion of 
one of the glassy colonies the growth at the point touched soon starts to 
become transparent or glassy, sometimes killing out all the micrococci and 
replacing these by fine granules. Experiments showed that the action is more 
rapid and complete with vigorous growing young cultures than with old ones, 
and there is very little action on dead cultures or on young cultures that have 
been killed by heating to 60°C. 

Anaerobia does not favour the action, although it appears best in tubes 
that are capped with gutta-percha tissue. The transparent growth when diluted 
(one in a million) with water or saline was found to pass the finest porcelain 
filters (Pasteur-Chamberland F. and B. and Doulton White) with ease, and 
one drop of the filtrate pipetted over an agar tube was sufficient to make that 
tube unsuitable for the growth of the micrococcus— that is, if the micrococcus 
was inoculated down the tube as a streak, this would start to grow, but would 
soon become dotted with transparent points which would rapidly extend over 
most of the growth. The number of points from which this starts depends 
upon the dilution of the transparent material, and in some cases it is so active 
that the growth is stopped and turned transparent almost directly it starts. 
This condition or disease of the micrococcus when transmitted to pure cultures 
of the micrococcus can be conveyed to fresh cultures for an indefinite number 
of generations; but the transparent material will not grow by itself on any 
medium. 

If in an infected tube small areas of micrococci are left— and this usually 
happens when the micrococcus has grown well before becoming infected— 
these areas will start to grow again and extend over the transparent portions, 
which shows that the action of the transparent material is stopped or hindered 
in an overgrown tube; but it is not dead, for if a minute portion be transferred 
to another young culture of the micrococcus it soon starts to dissolve up the 
micrococci again. Although the transparent material shows no evidence of 
growth when placed on a fresh agar tube without micrococci, it will retain 
its power of activity for over six months. It also retains its activity when made 
into an emulsion and heated to 52° C, but when heated to 60° C. for an hour 
it appears to be destroyed. It has some action, but very much less on 
Staphylococcus aureus and alhus isolated from boils of man, and it appears to 
have no action on members of the coU group or on streptococci, tuberc'e bacilli, 
yeasts, etc. The transparent material was inoculated into various animals, and 
was rubbed into the scratched skin of guinea-pigs, rabbits, a calf, a monkey, 
and a man, bat all the results were negative. 

When continuing my investigation of infantile diarrhoea and vomiting, for 



15 



Aug. 19, 1922] SECTION OF MICROBIOLOGY. [MJrcALjo'uTKAL 

the Local Government Board in 1915, similar experiments were carried out with 
material obtained from the intestinal tract. After certain difficulties had been 
overcome it was found that in the upper third of the intestine, which contained 
numerous bacilli of the colon group, some larger forms were also present. In 
some cases these grew in far greater number than the ordinary types of bac- 
teria, but this was only so when precautions were taken to eliminate the action 
of a dissolving substance which infected the colonies so rapidly that they were 
dissolved before attaining a size visible to the eye. Here, then, was a similar 
condition to that found in vaccinia, and the greatest difficulty was experienced 
in obtaining the bacilli free from the transparent dissolving material, so rapidly 
was the infection increased and carried from one colony to another. At first 
this bacillus was not believed to be a member of the dysentery-typhoid-coli 
group, but I have now no doubt that it was a special large form of a member 
of this group. Unfortunately 1 lost my cultures during the war before all the 
characters of the bacillus were completely worked out. 

Similar though not such definite results were also obtained with a micro- 
coccus and a member of the coli-typhoid group of bacilli which were obtained 
from the intestinal mucous membrane of a dog sufiFering from acute distemper, 
and I obtained some evidence that the difficulty often experienced in isolating 
certain known pathogenic micro-organisms might be due to the same cause. 
In my paper I also pointed out that results similar to those obtained with 
vaccinia and with bacilli from the intestinal tract would probably be obtained 
in cases of true dysentery; but I was unable at the time to investigate dysen- 
tery, as my scheme for research on this disease which was submitted to the 
War Office in 1914 was not accepted. Shortly after the publication of my 
work, however, I went to Salonica in charge of the base laboratory. While 
there in 1916 the subject was again discussed on several occasions with the 
English, French, and Canadian bacteriologists; but at the time the phenom- 
enon was not accepted as one of much importance, and my fresh proposals 
for continuing the work on dysentery failed to meet with approval. As, how- 
ever, I had predicted, a similar condition was later found to occur in cultures 
of true dysentery bacilli, particularly by Dr. d'Herelle, who carried out a 
series of important experiments on these bacilli. Dr. d'Herelle's published 
researches appear to me to confirm in the main my own results with micrococci 
and members of the coliform group of bacilli. 

I will now pass on to consider the conclusions I drew from mv experiments. 
In the case of vaccinia it is clear that the transparent material contains an 
enzyme, and this is destroyed on heating to 60° C. for one hour. It also 
increases in quantity when placed on an agar tube containing the micrococcus, 
and this can be carried on indefinitely from generation to generation. If it is 
part of the micrococcus it may be either a stage in its life-history which will 
not grow on ordinary media but stimulates fresh cultures of the micrococcus 
to pass into the same stage, or an enzyme secreted by the micrococcus which 
leads to its own destruction and the production of more enzyme. The fact that 



16 



Aug. 19, 1922] THE BACTERIOPHAGE. [ MEorcAfjL'jRNAL 

the transparent portion cannot be grown except on the micrococcus makes 
it impossible to obtain any definite evidence on these points. There is this, 
however, against the idea of a separate form of hfe: if the white micrococcus 
is repeatedly plated out a colony from the last plates may give a good white 
growth for months when subcultured at intervals on fresh tubes; eventually, 
however, most pure strains show a transparent spot, and from this the trans- 
parent material can be obtained once again. Of course, it may be that the 
micrococcus was never quite free from the transparent material, or this may 
have passed through the cotton-wool plug and contaminated the micrococcus, 
but it seems much more probable that the material was produced by the micro- 
coccus. Incidentally, this apparent spontaneous production of a self-destroying 
material which, when started, increases in quantity, may be of interest in 
connexion with cancers. In any case, whatever explanation is accepted, I do 
not think my experiments definitely disproved the possibility of its being an 
ultramicroscopic virus, because we do not know for certain the nature of such 
a virus. If the transparent portion were a separate virus, it might be vaccinia, 
or it might be some contaminating non-pathogenic ultramicroscopic virus, for 
it is conceivable that whereas a non-pathogenic variety might grow on micro- 
cocci or bacilli, a pathogenic variety might grow only in the animal it infects. 
As the animal experiments were negative, there is no evidence that it is 
vaccinia, although such a virus might lose its virulence when grown outside 
the body. On the other hand, no evidence was obtained that it was a non- 
pathogenic variety. On the whole it seems probable, though by no means 
certain, that the active lytic material is produced by the micrococcus, and 
since it leads to its own destruction and can be transmitted to fresh healthy 
cultures, it might almost be considered as an acute infectious disease of micro- 
cocci. 

I have now described my original experiments and conclusions, and I will 
pass on to consider the views of other workers in this field. Where differences 
of opinion exist the controversy has centred chiefly round the experiments 
that have been carried out to determine the source and nature of the lytit 
material. I have already pointed out that when this material is diluted and 
filtered and pipetted on to a tube of medium just before inoculating the tube 
with the micro-organism to be lysed, then the action starts at definite points 
throughout the growth. Dr. d'Herelle obtained a similar result in his experi- 
ments with dysentery bacilli, and he considers this a strong point in favour 
of the view that the lytic material is a separate living micro-organism. I can- 
not, however, for several reasons, agree that such experiments prove this. First, 
it is well known, notably in the case of certain starches, that the diastatic 
enzymes do not in every case dissolve up the grains evenly, but start at certain 
points, causing pitting and erosion of the grains; at the same time some grains 
are more susceptible to the enzyme than others. In this case no one has sug- 
gested that the imevenness of the action proves the diastatic enzyme to contain 
a living ultramicroscopic virus. I have noted also with bacteria that not only 



17 



Aug. 19, 1922] SECTION OF MICROBIOLOGY. [ MEmcAL Journal 

does the lytic agent attack some members of a culture before it attacks others, 
but individual members may be pitted and eroded, and particularly is this so 
with certain large forms of dysentery bacilli which I shall consider later. Again, 
it is well known in the case of all bacteria that certain members of a culture 
are more resistant to chemicals and to specific lysins produced in animals, and 
it may well be that the lytic action on the bacteria starts from a number of 
distinct points because these points happen to contain specially susceptible 
micro-organisms, and it is only when the action is started and the lytic sub- 
stance increases in quantity and in concentration that the more resistant 
members become lysed. 

Then there are the very interesting experiments of Professor Bordet and Dr. 
Ciuca. These workers found that when a lysin was produced by an animal 
against a coliform bacillus, this lysin not only dissolved up fresh cultures of the 
coliform bacillus, but the lytic effect could be transmitted from culture to cul- 
ture. These results certainly do not favour the view that the lytic agent is a 
definite living micro-organism. Moreover, there are my own original experi- 
ments where, after obtaining normal growths of micrococci for a number of 
generations, eventually some fresh subcultures started to become lysed, 
apparently spontaneously. But these experiments appear to me to be evidence 
not only against the view that the lytic material is a definite living organism, 
but also against the view of Bordet and Ciuca that the lytic agent arises from 
an association of the bacterium with cells from the animal body. d'Herelle 
has also obtained the lytic agent by associating bacteria with filtrates of soil, 
etc., and I agree with him that these experiments do not support Bordet's 
view, although at the same time I do not think they support the view that the 
lytic agent is a definite living micro-organism. The apparent spontaneous pro- 
duction of the lytic agent in some of my pure cultures is, I think, evidence 
against both the views mentioned. It may, of course, be argued, as I men- 
tioned at the time, that my cultures were never really quite free from the lytic 
agent, but if this view is accepted regarding my repeatedly plated cultures, 
then it is reasonable to suggest the same regarding the cultures used by 
d'Herelle and by Bordet. 

Recent experiments, in fact, have in no way changed my views, and I repeat 
my original opinion regarding the lytic agent of the micrococcus— namely, that 
"the possibility of its being an ultramicroscopic virus has not been definitely 
disproved," and that "it seems probable, though by no means certain, that the 
active transparent material is produced by the micrococcus"; and I hold the 
same view regarding the lytic agent which various workers and myself have 
found associated with the dysentery-typhoid-coli group of bacilli. 

However, as I have already pointed out, it is just possible that an ultra- 
microscopic virus may be of the nature of an enzym.e, and if so, the original 
source of such a virus might be the cell it infects: in remote ages possibly a 
normal enzyme which has gradually developed to take on a pathological action 
as it has passed through an infinite number of generations of cells, either 



18 



Aug. 19, 1922] THE BACTERIOPHAGE. [muo'.cal Jouknai. 

bacterial, vegetable, or animal, according to its source, and which possesses the 
powei either of directly increasing in quantity or of stimulating the cell to 
produce more pathological enzyme. On the other hand, it is conceivable that 
if this lytic agent be a pathological enzyme, it might bear the same relation 
to the normal enzyme that the cancer cell does to the normal cell. These 
possibilities may sound very improbable, but at the same time some such 
explanation would account for the extremely specific nature of many of the 
ultramicroscopic viruses, and it would certainly explain the absence of 
visibility and of growth on artificial media. After all, excluding such organisms 
as the lepto-spirillum of yellow fever, which probably belongs to an entirely 
different group, there are fundamental differences between the smallest micro- 
organisms known and the various ultramicroscopic viruses. 

Cultivations of the minutest micro-organisms known show all the charac- 
teristics of ordinary bacteria. If these are obtained from the soil many of them 
grow well. On the other hand, if ultramicroscopic viruses are simply more 
minute members of the group of bacteria, there must be thousands of wild 
varieties in the soil, and yet no one has succeeded in obtaining a definite 
growth of a single variety on any solid medium. Minuteness of size might 
account for their being invisible, but it will not explain the absence of visible 
growth on artificial solid media. It is true that certain workers at one time 
claimed to have obtained growths of pathogenic varieties in Noguchi's medium, 
but in the absence of definite confirmation these experiments need not be 
considered here. 

Professor Bordet and Dr. Ciuca have carried out some most important 
experiments with the dysentery-typhoid-coli group of bacilli on the specificity 
of the lytic material, and have obtained results which indicate not only that the 
lytic agent can be made to break down allied bacilli, but also that certain 
resisting and otherwise changed strains can be obtained. As I have already 
mentioned, my micrococcus lysin had little effect on Staphylococcus aureus; 
but I did not carry out many experiments of this nature after obtaining the 
apparent spontaneous production of the lytic agent in my pure cultures of 
micrococci, as this result appeared to me to make it difficult to draw definite 
conclusions from such experiments. Many other workers, but particularly Dr. 
d'Herelle and Dr. Andre Gratia, have also carried out important experiments 
in this branch of the subject, but I must leave other speakers to deal with it, 
as it is impossible in the space at my disposal to do justice to their work. 

There is, however, another aspect of the subject which I should like to 
mention. As is well known, in pure cultures of such bacilli as dysentery, 
typhoid, and coli, one sometimes meets with forms which are considerably 
larger and longer than the average bacillus, and these may be found in patho- 
logical material containing these bacilli, being not uncommon in urine in cases 
of cvstitis. Hort and others have described these large forms in considerable 
detail. 

In certain experiments dealing with the lytic material I observed these long 



19 



Aug. 19, 1922] SECTION OF MICROBIOLOGY. [Mnll'.f JourLl 

forms in greater number than usual, and it was thought that they might in 
some way be connected with the lytic agent associated with this group of 
bacilli. There was also the possibility that they might be distinct but symbiotic 
bacilli growing only with the dysentery or other bacilli forming the culture; 
they might be mutations, or, again, they might be special forms with special 
work to do, like bees in a swarm, and, alone or in association with the lytic 
agent, might prove to be of importance in connexion with the pathogenicity 
of cultures and the production of immunity in the host. 

The work was interrupted by the war, but eventually the results were 
published in the British Journal of Experimental Medicine (October, 1920), 
and I shall do no more now than consider their possible relation to the lytic 
agent. In the first place, it must be noted that these large forms occur much 
more frequently in pathological material than they do in cultures on artificial 
media. In cases of infantile diarrhoea and vomiting I have found them in large 
number associated with a lytic agent in the upper part of the intestinal tract, 
and it is only after the action of the lytic agent is eliminated that they can be 
easily cultivated. The nature of the bacilli from such conditions is now being 
investigated. 

Most of my experiments were carried out with the Shiga type of dysentery 
bacillus. The first point of importance to be noted is that the large or "special 
forms" occur in all pure cultures of bacilli belonging to the dysentery-typhoid- 
coli group, besides occurring in cultures of the influenza bacillus and other 
micro-organisms. Further, they are much more numerous in very young cul- 
tures than in older ones, and practically disappear in cultures that are twenty- 
four hours old. I observed also, in cultures that had been growing for six to 
twelve hours, that these "special forms" were often partially dissolved. From 
these early experiments I concluded that the bacilli normally produce "special 
forms," and that these, when presumably of no use to the bacterial community 
of a pure culture, are dissolved by a lytic agent which is also present, and that 
this lysin prevents the special forms from multiplying, and interferes with 
their isolation as growths free from the normal small bacilli. I found, however, 
two methods by which these "special forms" could be isolated. They may be 
obtained either by repeatedly plating out on litmus-maltose agar tubes, or by 
growing in an emulsion of dead coliform, typhoid, or some similar bacillus, and 
then plating out on the maltose agar medium. Three fairly distinct and stable 
types were isolated, but it will be unnecessary to consider these in detail. The 
chief points which I wish to bring to your notice are these: The large bacilli 
proved to be "special forms" of the bacterium from which they were obtained, 
and were easily agglutinated by the specific serums. I also obtained some 
evidence that they were more pathogenic when produced by the normal bacilli 
than after repeated multiplication by division. When isolated and grown they 
became more resistant to the lytic agent. In most of the cultures numerous free 
granules were present, while many of the bacilli contained similar looking 
granules in the fusiform and round swelling which they presented. 



20 



Aug. 19, 19221 THE BACTERIOPHAGE. [ McIrcAf JoukLl 

These researches seem to indicate another possible explanation of the lytic 
action of the substance under review. If this substance is, after all, not patho- 
logical to the bacterium, but is a normal product of its activity, then it must 
be produced for some special purpose w^hich is advantageous to the life of 
that variety of bacterium. If the "special forms" of bacilli which are found in 
normal cultures and are so soon dissolved up are produced for the benefit of 
the bacterial community of a pure culture, then it may be that the lytic agent 
is formed for the purpose of carrying out this lytic action, and possibly setting 
free toxins contained in the special forms. Or it may be that it sets free "anti- 
toxins," or substances that will neutralize the toxic substances of the host it 
infects, or such substances as are produced by other varieties of bacteria with 
which it struggles for existence when outside the animal body. Probably all 
bacteria produce substances of a toxic nature, but the possibility of their pro- 
ducing antibodies which will neutralize the toxins of other varieties is one that 
has interested me for some time. My experiments in this direction lead me to 
believe that this may be the case, but a research of this nature is rather difficult 
to carry out technically, and I shall do many more experiments before com- 
mitting myself to this view. If, however, additional experiments should con- 
firm my opinion, and if the special forms play some part in this process, one 
can quite understand that the special forms would be useless among the mem- 
bers of a pure culture growing on artificial media. In fact, they might be 
directly detrimental to the bacterial community, in that they would use up 
the food supply and overgrow the other bacilli which are necessary for a con- 
tinuation of the life of that species or variety. In this case the lytic agent would 
perform a good purpose in getting rid of the useless special forms. It may be 
argued that this will not explain the lysis of the normal forms of bacilli, but 
then the conditions of cultivation on artificial media are not those of nature. 
In fact, in the process of evolution the bacterium did not develop its characters 
under the influence of cultivation in incubators, and an action that might be 
restrained to the limits of useful purposes in nature might very well extend 
beyond those limits under such an abnormal environment as that presented 
by an uncontaminated tube of agar. 

Some of my views and the possibilities I have suggested are no doubt open 
to criticism, but I claim that the discovery of the filter-passing lytic agent in 
association with bacteria offers a large field for research, and I suggest that 
this field has been further extended by the isolation of "special forms" of bac- 
teria and by the possibility of demonstrating the production of bacterial anti- 
toxins for the neutralization of toxins produced by other varieties of bacteria. 
Moreover, it must be remembered that all the vital processes of a bacterium 
have some relation to each other, and it is therefore necessary to study these 
problems together, or the true significance of any one may not be fully appre- 
ciated. 



21 



Aug. 19, 1922] SECTION OF MICROBIOLOGY. [M.IrcA?j"JKNAi. 

III.-CONCERNING THE THEORIES OF THE 
SO-CALLED "BACTERIOPHAGE." 

By J. BoRDET, M.D., 

Director of the Pasteur Institute, Brussels. 

[The following explanatory statement was read by Dr. Gratia for Professor 
Bordet.] 

Through Dr. Gratia I obtained access to the paper Dr. d'Herelle intends to 
present on the lytic phenomenon due to the so-called "bacteriophage." I was 
not a little surprised to find that Dr. d'Herelle in this paper attributes to my 
co-worker, Dr. Ciuca, and to myself, as regards the intimate nature of this 
phenomenon, an opinion which is wholly different from what we felt entitled 
to uphold from the very beginning of our studies on the subject.^ Dr. d'Herelle 
quotes our names next to Kabeshima's, and enlists us among the authors who 
assume that the lytic principle is a leucocytic secretion. In fact, this view 
seems to us altogether untenable, and is almost the reverse of the opinion we 
have constantly emphasized. 

I think we were first to advocate the view that the lytic principle is pro- 
duced by the microbe itself which shows the lysis— in other words, that the 
transmissible lysis is in reality an autolysis betraying a nutritive vitiation 
primarily started by external influences, an example of which may be the con- 
tact with a leucocytic exudate. No doubt it would be quite unnecessary to 
translate literally the many passages of our papers where this assumption is 
advocated. Some lines, however, may be quoted: 

"External influences such as that of a leucocytic exudate modify the bacterium, 
inducing the latter to elaborate a lytic substance capable of difl:using itself and bring- 
ing about the same autolytic phenomenon through successive cultures. When the 
autolytic process occurs a large number of the microbes present may perish, but some 
of them, being more resistant, are, during a certain length of time, still capable of 
reproduction in spite of their producing the active principle, thus imparting to new 
cultures of the same microbe the same autolytic tendency." 

In another paper we add: 

"According to d'Herelle, the lysis is due to a living being, to a filtering virus. We, 
on the contrary, believe that the lytic principle originates from the bacteria them- 
selves, which, when touched by this active substance, are capable of regenerating it, 
the factor responsil^le for the phenomenon being thus unceasingly reproduced— on 
the condition, however, that the bacteria be still living and provided with the 
alimentary substances necessary to their growth." 

I wonder how Dr. d'Herelle could possibly give such an erroneous account 
of our work as in his paper. The many authors who have written on the 
subject did not, like Dr. d'Herelle, misinterpret a theory which we have so often 
and so distinctly outlined and explained. I shall allow myself to quote, for 
instance, the paper recently published by Dr. Bruynoghe," who writes: 

"According to Bordet and Ciuca, the microbes undergo— throuffh the agency of 
a leucocytic exudate— a modification by which they are henceforth capable of 
elaborating an autolytic principle, this property being further transmitted to the 



22 



Aug. 19, 1922] THE BACTERIOPHAGE. [ M.Irc.f JouTk.l 

following generations by the germs which were sufficiently resistant, and thus could 
multiply. This interesting view permits the understanding of the fact that the lytic 
principle is only regenerated when the bacteria are living, since the theory asserts 
that this principle is produced by the bacteria themselves." 

I think there is no need to dwell longer on the subject. But one must agree 
that I could not refrain from correcting d'Herelle as regards our views, nor 
from presenting them again as they are expressed in all of our papers. The 
mere titles of these are clear enough; we always designate the phenomenon 
under the name of "the microbian transmissible autolysis." 

References. 
iC.R. Soc. Biologie, October, 1920. -Le Scalpel, Mtn-ch. 1922. 



IV.-ANDRE GRATIA, M.D. 

Pasteur Institute, Brussels. 



1. The Twort phenomenon and the d'Herelle phenomenon are identical. 
They are two different aspects of one and the same phenomenon: the trans- 
missible lysis of bacteria. 

When the "dissolving material" of Twort found in diseased agar cultures 
of micrococci obtained from vaccinia lymph is transplanted into a young broth 
culture of staphylococci a dissolution of the latter occurs, and the filtrate of 
the dissolved culture exhibits all the characteristics of a typical staphylococcus 
bacteriophage according to the definition of d'Herelle. 

On the other hand, typical staphylococcus bacteriophage could be obtained 
also by other means— namely, by the leucocytic exudate technique of Bordet 
and Ciuca, or by the puncture of a subcutaneous abscess. When small 
amounts of this staphylococcus lytic agent are introduced in melted agar which 
is afterwards slanted and seeded with sensitive staphylococci a culture results, 
apparently normal at the beginning, but which, a little later, turns into the 
typical glassy transparent material of Twort. In other words, the Twort 
phenomenon leads to the d'Herelle phenomenon, and, inversely, the d'Herelle 
phenomenon leads to the Twort phenomenon. 

2. There are no unquestionable proofs that the bacteriophage is a living 
organism. 

The assumption of the bacteriophage being a filtrable virus for bacteria 
was suggested by two main facts: (a) The power of reproduction possessed 
by the lytic agent, and ( b ) the localization of the lysis to certain round spots 
of clarification when a very diluted lytic agent is poured over the surface of 
an agar culture of sensitive bacteria. Although easily explained by the virus 
theory, yet both facts are not unquestionable proofs of the living nature of 
the bacteriophage, because they are by no means exclusive features of living 
beings. 

Fire is not living, and yet fire is endowed with power of reproduction. 
When once lighted, thanks to an initial impulsion such as an electric spark or 



23 



Aug. 19, 1922] SECTION OF MICROBIOLOGY. [ MEorcAL jL"rnal 

the mere striking of a match, it can be indefinitely reproduced if fuel is pro- 
vided. A still more striking, because more biological, example is found in blood 
coagulation. Suppose a series of tests tubes containing a stable plasma— bird's 
plasma, for instance— which will remain indefinitely fluid. To the first tube 
we add just a few cubic centimetres of distilled water. As a result of that 
initial thromboplastic action, which does not need to be repeated to the future, 
thrombin suddenly appears in the first tube and the plasma clots. If a few 
drops of the exudate serum in the first tube are pipetted off and poured in the 
next tube, this second tube clots, in its turn, with a new regeneration of throm- 
bin, which, transferred in the third tube, brings about the coagulation of that 
tube with again a new production of thrombin, and so on indefinitely. In this 
way we realize the transmissible coagulation of blood in series, with the con- 
tinuous regeneration of thrombin, and thrombin is not a living being. 

The localization of the lytic action of diluted bacteriophage can be 
explained by the hypothesis of a chemical substance as well. It must be kept 
in mind that a culture is not a homogeneous whole, but made up of organisms 
showing all kinds of qualitative and quantitative individual differences— that 
is, as far as their susceptibility to the lytic agent is concerned. When a very 
concentrated lytic agent is poured over the surface of an agar culture an almost 
complete dissolution occurs, with the exception of just a few organisms 
resistant enough to overwhelm the strong action of the concentrated lytic 
agent. On the other hand, when a diluted lytic agent is used only the few 
extremely sensitive bacteria will be influenced, and each of them becomes a 
centre of regeneration of the lytic agent, which, diffusing evenly in every 
direction, produces perfectly round spots of clarification very often surrounded 
by a kind of halo of diffusion. Between these two extreme conditions all kinds 
of intermediate degrees exist. Further, any substance, living or not, is com- 
posed of particles, molecules, atoms, or ions. When we pour out a glass of 
soda water, there appear on the wall of the glass small round bubbles of gas, 
the size of which increases exactly as the so-called colonies of bacteriophage, 
and yet gas is not a virus. 

3. The idea of the bacteriophage being a product of bacterial activity is 
suggested by the close parallelism existing between the regeneration of the 
lytic agent on the one hand, and the activity of growth of the bacteria on the 
other hand. 

No regeneration ever occurs in dead cultures, nor in living cultures when 
put in such conditions that they cannot grow— in saline emulsions of bacteria, 
for instance, or at low temperature. A slight lysis, with but a small regeneration 
of lytic agent, is induced in the slow-growing culture of B. coli in a svn- 
thetic medium. On the contrary, an abundant regeneration occurs in a fast- 
growing culture in broth. A recently seeded broth culture to which is added 
just a trace of lytic agent will not be inhibited; but a few hours later, at the 
very moment the culture reaches its acme of growth, a rapid dissolution 
occurs with an abundant regeneration of lytic agent. 



24 



Aug. 19, 1922] THE BACTERIOPHAGE. [McIrcAf JoukLl 

4. The conception of the bacteriophage being a chemical substance is 
favoured by the chemical-Hke affinity existing between a given lytic agent and 
the corresponding susceptible strain. 

I first observed that small amounts of lytic agents lose a certain part of their 
activity when put together with too thick emulsions of sensitive bacteria. 
Bordet, with a different technique, could even obtain the complete disappear- 
ance of traces of lytic agent in the same condition. Still more convincing are 
the results of Yaumain and of Da Costa, who observed the absorption of rela- 
tively important amount of lytic agent by dead emulsions of the corresponding 
sensitive bacteria. This specific affinity which is the necessary condition for 
a lytic agent for inducing the dissolution of a given bacterium is not favourable 
to the virus theory, because we question how a virus could be definitely fixed 
by dead bacilli, which, however, it is unable to attack. 

5. The bacteriophage is not one and the same antigen. Several lytic agents 
showing antigenic specificity must be considered. 

The coli lytic agent can be completely neutralized by proper amounts of 
corresponding coli antilytic serum, but is not at all affected by staphylococcus 
antilytic serum, which, on the other hand, is only able to neutralize staphylo- 
coccus lytic agent and not coli lytic agent. This neutralization reaction is thus 
specific, and demonstiates the plurality of the bacteriophage. 

The non-specific results obtained with the alexin fixation reaction and 
advocated by d'Herelle in favour of the unicity of the bacteriophage, are of no 
value, because they are vitiated, as can be easily demonstrated, by the 
presence in the bacteriophage of bacterial dissolution products which have lost 
their specificity and play therefore the role of common antigen between 
different lytic agents. 



25 



Adsorption of Bacteriophages to Homologous Bacteria 

II. Quantitative Investigations of Adsorption Velocity and Saturation. 
Estimation of the Particle Size of the Bacteriophage 

By 

M. SCHLESINGER 

Institute for Colloid Research, Frankfurt A.M. 



It was shown in the preceding communication that bacteriophages of a lysate 
do not represent a uniform population; instead they are composed of several 
classes of widely varying adsorbabilities. The following equation expressing the 
kinetics of adsorption of phages to homologous bacteria was derived : 

71 

2.3 log - = kht , (1) 

nt 

where no represents the initial concentration of free phage, n« the concentration 
after a reaction time t and h the concentration of bacteria per ml. However, this 
equation is applicable only to experimental conditions under which the observed 
change in the concentration of free bacteriophages is mamly due to the adsorption 
of a single class of particles of uniform adsorbability. Lysates of the coli 88 bac- 
teriophage employed in our experiments were found to consist almost entirely of a 
uniform class of rapidly and irreversibly adsorbable phages when conditions were 
such that the concentration of sensitive bacteria and the reaction time was varied 
only within limits that allowed the decrease in concentration of free phage not to 
exceed 90 to 95%. 

It shall be shown in this communication that under these conditions equation 
(1) is so well satisfied that wide variations of the different experimental param- 
eters produce no change in the value of the velocity constant k calculated from 
the experimental results. After establishing the value of k for adsorption to dead 
as well as to living bacteria, experiments will be presented that are designed to 
determine the bacterial "saturation capacity", i.e. the number of phage particles 
that a single bacterial cell can adsorb. Finally, an attempt will be made to calculate 
the particle size of the phage from the adsorption velocity constant. Further 
insight shall be gained into the adsorption mechanism by a comparison of the 
values so calculated with the results of direct determinations of the particle size. 

Translated from the German and reprinted by permission of 

the Springer Verlag from Zeitschrift fur Hygenie und 

Immunitaetsforschung, 114, 149-160 (1932). 

26 



Adsorption of Bacteriophages to Homologous Bacteria 

Measurement of the Adsorption Velocity. 

In order to test equation (1), we shall determine, as a function of time, the in- 
fectivity of mixtures of known concentration of heat-killed bacteria and input 
bacteriophages, so that the value of the expression 

2 3 no 
k = -r- los; — • 
bt ^ nt 

whose constancy is demanded by equation (1), can be calculated. 

These experiments are carried out in the following way. A heat-killed bac- 
terial suspension of known concentration is diluted into broth, so that after ad- 
dition of the phage the concentration of bacteria in the reaction mixture is 6. 
After this dilution has been warmed to 37°, a measured amount of a lysate di- 
lution of exactly known phage titer is added and the time noted. In order to keep 
the temperature constant, the reaction mixture is placed in an incubator, in the 
case of long experiments, or, otherwise, in a water bath. After various times have 
elapsed, a measured aliquot of the reaction mixture is quickly diluted 10 or 100- 
fold, so that for all practical purposes, the reaction is stopped ^ In this manner, 
reaction times down to 2 minutes can be investigated, after some practice. For 
reaction times of more than 1 hour, in cases where the anticipated titer permits 
this procedure, the sample is plated directly on agar plates. Under the present 
experimental conditions it is unnecessary to separate the dead bacteria from the 
reaction mixture before plating, since only irreversible adsorption is involved. 

The detailed results of one experiment, and the calculations based thereon, 
shall be presented here as an illustration. 0.25 ml of the heat-killed broth culture 
of coli bacteria, containing 4 X 10* bacteria ml, was added to 4.65 ml broth. 
After warming the bacterial suspension to 37°, 0.1 ml of a thousand-fold dilution 
of a lysate of titer 8.5 X 10* was added to it. The mixture was placed in the incu- 
bator; after 1, 2, and 3 hours, 0.1 ml aliquots were removed from the mixture and 
spread at once on agar plates. The initial titer of the mixture, n<„ was 1.7 X 10*; 
the titers, after 1, 2, and 3 hours were found to be 5.8 X 10^ 2.4 X 10^ and 7.2 
X 10^ With the known value of 6 = 2.0 X 10^ and the time measured in 
seconds, k can be calculated. 

After 1 hour 

2^3 Tio ^ 2.3 1.7 X 10^ 

ht ^^ nt 2 X 10^ X 3600 ^^ 5.8 X 10 » 



After 2 hours 
After 3 hours 



= 3.2 X 10- 11 log 2.9 = 1.5 X 10-" 



3 2 X 10- " 
J. _ '^- -r ^^ log 7.1 = 1.4 X 10-" 



3 9 V 10"" 
k = •^- ^^^ log 23.6 = 1.5 X 10-" 



^A 100-fold dilution reduces the reaction velocity by a factor of 100. 



27 



M. Schlesinger 

The values of k (multiplied by 10^ '), calculated on the basis of our most exten- 
sive experiments, are summarized in Table I. 

For these experiments, a phage lysate of titer 6.5 X 10* and a heat-killed bac- 
terial suspension (70° for 1 hour) were used. Prior to heating, the bacterial 
suspension was grown from a slant agar culture to a density of 5.8 X 10^, as de- 
termined by colony count. However, the value of 6.5 X 10*^ used in the calcula- 
tions was determined by microscopic cell count after heat-killing. 



Table I 







6 = 1.6 X 108 


b = 6.5 X 10^ 


h = 3.3 


X 10^ 












to 


lo 


^ 


-1- 


CO 


_^ 


CO 


CO 


M 




QO O 


o 


o 


^ 


o 


o 


o 


Ss ° 


« o 




O -H 






_i 








O -H 


O -H 


c ^ 




















.a T3 


xX 


X 


X 


X 


X 


X 


X 


;x 


xX 


a 8 


.^ ^ 


lO 


iC 


lO 


>c 


lO 


lO 


lO 


CO '^ 


-H aj 


"^ o 


c^ 


CO 


o 


o 


o 


d 


"^ CD 


^ «= 


H OQ 


II \ 


II 


II 


II 


II 


II 


II 


^ ". 


• ". 
























-o S 


S 


s 


c: 


i; 


'^ 


?i 


-o S 


-o S 


120 


1.1 




















240 


1 


2 


























360 


1 


1 


1 


5 


1.6 




















720 






1 


2 


1.3 




















1080 






1 


1 


1.1 




















1440 






1 


2 


1.4 




















1800 












1 





1.2 














3600 












1 


2 


1.3 


1.1 


1.2 


1 


7 






7200 


















1.2 


1.2 


1 


3 






10800 


















0.9 


1.0 


1 


5 






14400 






















1 


3 


1 


1 


28800 


























1 


3 


43200 


























1 


5 



It is apparent that, neglecting variations due to experimental error, the value of 
k remains virtually constant, in spite of the fact that in the different experiments 
the initial titer of phage has been varied by a factor of 100, the concentration of 
bacteria by a factor of 200 and the reaction time by a factor 360. The average 
value of the velocity constant in these experiments is 1.2 X 10~^^ 

In five additional, less extensive, experiments, carried out with cultures of coli 
88 grown either in broth or on agar, the following values of k were found: 1.0 X 
10-11, 1.4 X 10-11, 1.6 X 10-11, 1.0 X 10-11, and 14 x IQ-n (average values of 
3, 7, 9, 4, and 7 determinations). These fluctuations are no greater than the 
uncertamty in the determination of the bacterial concentrations 2. The average 
value of the velocity constant k, computed on the basis of 58 single determi- 
nations for the binding to dead bacteria, was 1.3 X 10- n. 



2As has already been mentioned, the heat-killing was carried out by heating to 70° for 1 hour. 
This temperature treatment has no important effect on the binding capacity of the bacteria. 
In a series of experiments, ahquots from a single broth culture of coli were heat-killed at 70° 
for i, 1, 2 and 3 hours; the adsorption velocity constants determined in e.xperiments utilizing 
these suspensions were 1.7 X lO-n, 1.6 x IQ-i!, 1.3 X lO"" and 1.4 X 10-". 



28 



Adsorption of Bacteriophages to Homologous Bacteria 

The velocity of adsorption of phage to living bacteria was measured either di- 
rectly, or by adding bacteriophages to a mixture of dead and living bacteria and 
then determining their distribution over the two components. 

A 6-hour-old agar slant of coli 88 is suspended in broth and freed of larger 
clumps by short centrif ugation ; half of this suspension is placed on ice, and the 
other half is heated for 30 minutes in a water bath at 70°. The cell concentration 
in each of the cultures is then determined by microscopic counts. 0.9 ml aliquots 
of each culture, as well as equal volumes of a two-fold dilution, are placed in the 
water bath at 37° and infected with 0.1 ml of a 1000-fold dilution of the phage 
lysate. After 3 and 6 minutes, or after 6 and 12 mmutes, aliquots of each mixture 
are diluted 10-fold and centrifuged at once at 8000 r.p.m. and the titer of the 
supernatant liquid determined. The values of k, calculated from the experi- 
mental results, are presented in Table II. (The uncentrifuged mixtures are also 
titrated; in the experiment with living bacteria, the infective titer remains equal 
to that of the original lysate ; whereas, in the experiment with dead bacteria, the 
surviving infectivity is always equal to that of the infectivity of the supernatant 
liquid.) 

The suspensions of living and dead bacteria are then mixed with one another 
m the following proportions: 5 -f 5, 3 + 7, 2 + 8, and 1+9. To each of 
0.9 ml of these mixtures, as well as to a suspension of only living bacteria, 0.1 ml 
of the lysate dilution is added and an aliquot of 0. 1 ml of each mixture spread on 
agar plates after 15 minutes. During this time no multiplication of the bacterio- 
phage takes place, whereas most of the phages are already adsorbed. Since any 
phage adsorbed to a living bacterium manifests itself by plaque formation, one 
may infer that any observed decrease in phage titer must represent phages ad- 
sorbed to dead bacteria. This decrease thus permits a calculation of the ratio of 
the adsorption velocity constants k (living) k (dead) which apparently indicates 
the distribution of phages over two competing components. The result of this 
experiment is presented in Table III. 

Microscopic counts of the suspension of dead bacteria indicated a concen- 
tration of 4.7 X 10^, of the suspension of living bacteria a count of 5.8 X lOVml. 
These figures are reduced by 10% in the reaction mixture through dilution upon 
addition of the phage lysate. 

The results of direct measurement of the adsorption velocity are presented in 
Table II. 

Table II 



Time in 


Heat-killed bacteria 


Living bacteria 


seconds 


6 = 4.8 X 10« 


6 = 2.4 X 108 


6 = 4.8 X 10« 


6 = 2.4 X 108 


180 
360 
720 


1.0 X 10-1' 
1.0 X 10-11 


1.0 X 10-11 
0.8 X 10-11 


2.4 X 10-11 
2.8 X 10-11 


2.7 X 10-11 
2.3 X 10-11 



Thus, according to this experiment, with the ratio of the two velocity constants 
being 2.6, the average value of k is 1.0 X 10~^^ for dead bacteria and 2.6 X 10~^^ 



29 



M. Schlesinger 

for living bacteria. The distribution of bacteriophages in a mixture of living and 
dead bacteria is presented in Table III. 

Table III 



In 10 parts of the mixture 


Plaques per 0.1 o.o. 


k (living) / k (dead) 


Living 
bacteria 


Heat-killed 
bacteria 


1000 
720 
420 
360 
280 




10 

5 
3 
2 

1 


5 

7 
8 
9 


2.6 
1.7 
2.2 
3.5 




Average 2.5 



The results of this method are thus in agreement with those obtained from the 
first method. Hence adsorption of phages to young living bacteria proceeds 2 to 
3 times more rapidly than adsorption to dead bacteria. If we multiply the value 
of k of 1.3 X 10~^\ determined in numerous experiments with dead bacteria, by 
the proportionality constant 2.6, we obtain the velocity constant of adsorption to 
living bacteria: k = 3.4 X 10~^^ 

Determination of the Saturation Capacity of Bacteria. 

The following method was selected for determining the maximum number of 
bacteriophages, referred to as the saturation capacity of the bacterium, that the 
average heat-killed bacterium can adsorb: samples are prepared which contain 
exactly the same number of dead bacteria but to which increasing initial concen- 
trations of bacteriophages are added. After a sufficient length of time has 
elapsed, all of the free bacteriophages capable of irreversible adsorption will have 
disappeared from each sample. The remaining phage titer will thus be independ- 
ent of the absolute value of the initial phage concentration, or, at any rate, will 
represent only a small fraction of it. In fact, the approximate constancy of the 
relative decrease in free phage should persist even when the allowed reaction time 
is insufficient for the attainment of the final state; for both parameters im- 
portant for the velocity of adsorption, the concentration of the bacteria and the 
relative proportion of phages of differing adsorbability, are constant for all of the 
samples. This picture will change, however, and change drastically, whenever 
the initial concentration of bacteriophages attains that multiple of the bacterial 
concentration which corresponds to the saturation capacity. For instance, when 
the initial concentration is only half of the saturation capacity, then the adsorp- 
tion velocity could be reduced, but only near the end of the adsorption process; 
when, however, the initial concentration of phage is twice that of the total 
bacterial capacity, then irrespective of the length of time of the experiment, only 
half of the initial phage population can be adsorbed. The saturation range thus 
manifests itself by a rapid increase in the ratio of final to initial phage titer in such 
an adsorption experiment. This ratio will vary from a value of about 0.01 prior 
to a saturating phage concentration to a value of about 1.0, once saturation has 



30 



Adsorption of Bacteriophages to Homologous Bacteria 

been attained. If no saturation is observable in such a series of experiments, even 
with the highest initial concentrations of phages obtainable, then the concentra- 
tion of bacteria must be decreased in a second series of experiments. In that 
event, of course, the reaction time must be correspondingly prolonged. 



Table IV. Bacterial Concentration 5.0 X 10''. Reaction Time: 1 Day. 



Initial Titer 


Final Titer | Ratio Phage /Bacteria | Relative Decrease 


5.0 X 108 


2.1 X 10« 


10:1 


1 :240 


2.5 X 108 


1.0 X 106 


5:1 


1:250 


1.0 X 108 


3.3 X 105 


2:1 


1:300 


5.0 X W 


1.1 X 105 


1:1 


1:450 


2.5 X 107 


1.8 X 105 


0.5:1 


1:140 


1.0 X 10^ 


3.1 X 10* 


0.2:1 


1 :320 


Table V. Bacterial Concentration 1.0 X 10^. Reaction Time: 4 Days 


9.0 X 108 


4.6 X 106 


90:1 


1 :200 


4.5 X 108 


2.0 X 106 


45:1 


1:220 


1.8 X 108 


5.0 X 105 


18:1 


1:360 


9.0 X 10^ 


2.6 X 105 


9:1 


1 :340 


Table VI. 


Bacterial Concentration 1.0 X 106. Reaction Time: 8 Days 




(Two samples became unsterile) 


1.1 X 109 


1.1 X 109 


1100:1 


1:1 


5.5 X 108 




550:1 




2.2 X 108 


8.4 X 107 


220:1 


1 :2.6 


1.1 X 108 




110:1 




Table VII. 


Bacterial Concentration 1.8 X 10*. Reaction Time: 8 Days 


7.6 X 108 


7.1 X 108 


420:1 


1:1.1 


3.8 X 108 


1.4 X 108 


210:1 


1:2.7 


1.5 X 108 


8.2 X 106 


84:1 


1:18 


7.6 X 10^ 


4.5 X 105 


,42:1 


1:170 


3.8 X W 


6.6 X 10* 


21:1 


1:570 


1.5 X 10^ 


2.7 X 10* 


8.4:1 


1 :550 



The experimental results are summarized in Tables IV to VII. The experi- 
ments were carried out with different dilutions of a three-hour broth culture of 
coH 88, heat-killed at 70°, containing 1.0 X 10» bacteria/ml (Table IV-VI) and 
with a dilution of a similar culture containing 1.8 X 10* bacteria ml (Table VII). 
As controls, samples or dilutions of the lysates were always placed in the incu- 
bator and their titers similarly determined at the conclusion of the experiments; 
these titers always agreed with the earlier assays of the lysates. 

It follows from the data of Table IV that the saturation capacity must be at 
least 10, while the data of Table V show that this capacity must exceed at least 
90. In the last two tables, the region of saturation has been attained, and their 
third and second lines, respectively, permit a more exact estimation. According 
to Table VI, 1.0 X 106 bacteria are saturated by (2.2 - 0.8) X 10« bacterio- 
phages. Hence, one bacterium is saturated by I40 phage particles. Similarly, it 
follows from the data of Table VII that the saturation capacity has the value of 130. 



31 



M. Schlesinger 

Therefore, a coli bacterium killed by heating to 70° is capable of irreversibly 
adsorbing an average maximum of 130 to 140 bacteriophages. 

Calculation of the Particle Size of the Bacteriophage from the Adsorp- 
tion Velocity and from the Saturation Capacity. Conclusions: 

a) Calculation of Particle Size from the Adsorption Velocity. The assumption 
was made in the derivation of equation (1), a formal description of the adsorption 
kinetics, that contact between one phage and one bacterium occurs entirely by 
chance and without benefit of any orienthig or attractive forces whatsoever. The 
collisions were thought to be mainly due to the Brownian movement of the 
particles^, to which, obviously the movement of the much smaller bacteriophages 
made the principal contribution. Since there exist well-known relations between 
Brownian movement, the diffusion coefficient, and the particle size, it is thus 
possible theoretically to correlate the latter with the adsorption velocity of the 
phage. It is, however, first necessary to make an assumption concerning the 
relation which might exist between the Brownian movement collision frequency 
of phages and bacteria and the number of adsorptive events which actually ensue 
as a consequence of these coHisions. 

The simplest assumption in this connection is that every collision leads to an 
irreversible fixation. If this assumption is valid, then the relation of the adsorp- 
tion velocity to the diffusion constant, D, of the phage follows at once from the 
formula on which M. v. Smoluchowsky has based his theory of the kinetics of 
coagulation^. This formula states that the quantity Jdt of a solute which diffuses 
in a time dt onto a sphere of radius R capable of fixing solute particles which touch 
it is 

Jdt = AirDRcdt 

where c is the concentration of the substance^. If we efjuate R to the radius of a 
supposedly spherical bacterial cell, substitute for c the number of free bacterio- 



^It is also possible that convection or mechanical agitation of the fluid could play a certain 
role. The error arising from neglecting the Brownian movement of the bacteria depends on the 
ratio of the bacterial diameter to that of the phage, and probably causes the estimate of the 
phage diameter to be too small by 10% in the following calculation. 

^Z. Phy.nk. Chem., 92, 140 (1917). An attempt — though insufficient — to apply the formula 
of the V. Smoluchowsky coagulation theory to bacteriophage adsorption has already been made 
by V. Angerer [Arch. f. Hyg., 92, .312 (1924)]. 

^The complete formula is 



Jdt = 4DRc 



R 

1 +-J^ 



dt 



where I indicates the time which has elapsed since the start of the experiment. As in the 
coagulation experiments of Zsigmondy analyzed bv v. Smoluchowsky, so also, under our 

R ' 

experimental conditions is the quantitv , small compared to imity, and can be neglected 

yJTrDt 

without significant error. Incidentally, Jdt represents in v. Smoluchowsky's application, as 

well as in ours, not the number of particles which actually diffuse onto the sphere in the time dt 

(after all, in our experiments there is only one bacteriophage for every hundredth or thousandth 

bacterial cell), but onh' the probability of an encounter of a particle with the adsorbing sphere. 



32 



Adsorption of Bacteriophages to Homologous Bacteria 

phage particles n, replace Jdt by the decrease —dn of the number of free phages 
which occurs in the time dt, and bear in mind that for the adsorption process there 
is available not one but simultaneously h adsorbing bacteria (referring our entire 
consideration to the unit of volume), then we can write 

— dn = ^TzDRhndt. 

If we compare this expression with the differential equation on which equation (1) 
was based 

— dn = kbndt 
then it follows that 

k = 4tDR or Z) = -A^ . 
■iirK 

the desired relation between adsorption velocity and diffusion constant of the 
bacteriophage. Now it will be recalled that the values of the velocity constant 
for adsorption to living and dead bacteria are not the same, since k for living 
bacteria is 2.6 times greater than k for dead bacteria. The reduced adsorption 
velocity for heat-killed bacteria evidently implies that a smaller fraction of the 
contacts leads to adsorption, since the collision frequency is the same whether the 
bacteria are living or dead. The assumption that every collision leads to fixation 
can therefore be excluded a priori for dead bacteria, while it could still be valid 
for living bacteria. In the latter case, the number of collisions is probably greater 
than that estimated since the motion of the coli cell was neglected in the calcu- 
lations. Nevertheless, the relatively slow motion of the bacteria is insufficient to 
explain the entire difference between the adsorption velocity to dead and to living 
bacteria, since the movement of the cells could hardly triple the collision fre- 
quency. Therefore, our comparison of the D values for living and dead 
bacteria will be based on our calculations of their respective k values. 

For the purpose of our calculations we shall equate R to the radius of that 
sphere whose surface is equal to the surface of a cylmder of 1.2 /x length and 0.5 m 
width, in which case R is equal to 4.3 X 10~^. The diffusion coefficient of the 
phage at the experimental temperature of 37° thus is 

^ 4 X 4.3 X 10-^ ^-^ ^ ^^ 

if the time is reckoned in seconds, as in the calculation of k. If this value is 
reduced to the unit of time generally used in diffusion experiments, i.e., the day, 
it follows that D is equal to 0.0055. On the basis of this value of the diffusion 
coefficient, one estimates from Einstein's formula^ a diameter of the bacterio- 



^This formula is 

CT 



'^ GirNnD 

where p is the particle radius, C the gas constant (8.32 X 10'' erg/degree), T the absolute 
temperature, A^ Avogadro's number (60.7 X 1022), and n the viscosity. It thus follows that 

8.32 X lOT (273 + 37) 



'^ Qw X 60.7 X 1022 X 0.007 X 6.3 X IQ-* 
or that the diameter is 102 m^. 



33 



= 5.1 X 106 cm 



M. Schlesinger 

phage of 100 m/x. On the other hand, utihzmg k for killed bacteria (1.3 X 10~ "), 
one calculates a particle diameter of 260 m/x. This latter value is undoubtedly 
too high; thus, in the case of heat-killed bacteria, it seems certain that not every 
collision can lead to fixation. It follows from this calculation that the adsorption 
velocity of the bacteriophage is not too great to be fully explained by Brownian 
movement alone. If one reverses this calculation and estimates the adsorption 
velocity constant from the known particle size of the bacteriophage (80 to 90 m/x; 
see below), one obtains a value of approximately 4 X 10~'^ It seems probable 
that this agreement in order of magnitude of different, independent estimates of 
the particle diameter can be taken as confirmation of the essential validity of 
the concepts on which these estimates were based. 

b) Calculation of the Particle Size from the Saturation Capacity. The particle 
size of the bacteriophage can be calculated from the saturation capacity, if it is 
assumed that the bacterial cell absorbs that maximum number of phages which 
can completely cover its surface with a monolayer of particles. The premise that 
the adsorption takes place in a single layer, i.e., that it results only from direct 
contact between phage and bacteria, is justified on the basis of the irreversible 
and specific nature of the adsorption process^. On the other hand, the assumption* 
that saturation occurs only after the entire surface of the bacterium has been 
covered by phages is inherent in the previous hypothesis that every contact 
between phage and bacterium leads to fixation. For every contact can lead to 
fixation only if the entire bacterial surface is capable of phage fixation^. 

In case of close-packing, 140 spheres of diameter 8 cover a surface of 140 5^ 
If this value is equal to the surface of the bacterial cell, that is to say 2.3 M^ then 
it follows that 5 = 127 m/x- If one considers, however, that random collisions of 
phage particles will hardly cover the bacterial surface in the closest packing 
assumed here, and that the number 140 refers to the adsorption capacity of 



''In the case of heat-killed bacteria, there can be no question of any penetration of the 
bacteriophage into the interior of the cell; at least there is no indication of such a possibility. 

^Obviously, this hypothesis refers to conditions under which there can be no question of any 
interference with adsorption by partial saturation; that is to say, under conditions of great 
excess of bacteria. The assumption that this hypothesis is not fully applicable to heat-killed 
bacteria will be considered below. 

^However, even if the entire surface of the bacterium were capable of phage fixation, it would 
not necessarily follow that every collision would lead to fixation ; other prerequisites of fixation 
might be appropriate direction and adequate intensity of the collision. In the event that only 
a number of isolated points, instead of the entire surface, is capable of fixation, then only those 
collisions which bring the phages into contact with these "valence points" could lead to fixation. 
The saturation capacity (S) then represents the number of such points. This assumption also 
makes possible an estimate of the phage diameter 5, on the basis of the equation 

This equation states that the adsorption velocity is proportional to the fraction of the bacterial 
surface capable of adsorbing the phage. The calculation, for which k for heat-killed bacteria 
must be employed since S refers to the adsorption capacity of heat-killed bacteria, leads to a 
value of 5 = 19 nxfi. If we assume the correctness of our direct determination of particle size 
of the phage (80-90 m^i) mentioned below, then this low estimate of the phage diameter indicates 
that almost the entire bacterial surface must be covered by "valence points." 



34 



Adsorption of Bacteriophages to Homologous Bacteria 

heat-killed bacteria, whose 2.6-fold reduction in adsorption velocity possibly 
reflects in inactivation of the phage receptors of a corresponding fraction of the 
cell surface, then one might revise this estimate of the phage diameter down to 
values of 61 to 120 m/x. 

Thus, two completely independent sets of data, the adsorption velocity and the 
saturation capacity, yield roughly the same value for the diameter of the bacterio- 
phage: 100 mM and 60-120 m^ respectively. These figures appear much too high 
when compared with most of the values for the particle diameter given in the 
hterature. In recent months, however, we have succeeded in measuring the 
sedimentation velocity in a centrifuge run at 7000 to 8000 r.p.m., of phage 
particles from lysates of the same coli phage selected for the present adsorption 
experiments. These measurements, which no doubt exceed all others in reliabil- 
ity, allowed us to calculate a particle size of 80 to 90 m/u. By means of a successive 
series of repetitive centrifugations, we were finally able to sediment into a pellet 
up to 99.99% of the active particles without encountering any decrease in the 
sedimentation velocity, i.e., without detecting any particles of smaller diameter, i" 

The agreement of the values of the phage diameter calculated indirectly from 
the adsorption velocity and from the saturation capacity, and the similarity of 
these values to the phage diameter determined directly, substantiates the validity 
of our assumptions. In any case, the formal kinetics of the process as well as the 
actual numerical values of the relevant parameters can be integrated, without 
contradiction, into the following simple picture. In each suspension of a mixture 
of bacteria and bacteriophages collisions occur between the individual particles of 
the two components. These collisions are entirely random, and their frequency is 
mainly determined by the Brownian movement of the bacteriophages. In the 
case of bacteriophage particles of maximum adsorption aflfinity (and in our coli 
lysate most of the particles are of this nature), the first such random collision 
between phage and bacterial cell leads to an irreversible fixation. More precisely, 
in the case of young, living bacteria almost every collision leads to fixation and in 
the case of heat-killed bacteria approximately every third collision does ' ^ In the 
case of phage particles of reduced affinity, only a fraction of the collisions leads to 
fixation; and here also the fixation is less stable and partially reversible. The 
bacteria conserve their ability to find additional bacteriophage particles as long as 
their surfaces are not completely covered by a monolayer of such particles. 
Investigations on the influence of the medium on the absorption process shall be 
communicated in a future paper. 

Summary 

1 . The equation describing the adsorption kinetics derived in a previous paper 

2.3 log - = kht 

rit 



i^The results of this investigation will be published in this journal. 

i^This statement refers only to the experimental conditions communicated here, that is to 
say, to a broth medium and to the temperature of the incubator. 



35 



M. Schle singer 

is satisfied for very wide variations in the parameters Uo, b, and t (initial concen- 
tration of phage, concentration of bacteria, and reaction time), provided that 
these variations remain in such limits that the adsorption velocity is still deter- 
mined by the same group of bacteriophages of maximal adsorptive affinity (i.e., 
Ut/no must not fall below 1/20). 

2. The adsorption velocity constant k has the value of 1.3 X 10~^^ for adsorp- 
tion of phage to heat-killed bacteria; k has an approximately 2.5 times greater 
value for phage adsorption to living bacteria. 

3. A heat-killed coli bacterium is capable of adsorbing, on the average, a 
maximum of 130 to 140 bacteriophage particles; this number is referred to as 
the saturation capacity. 

4. On the basis of the simplest assumptions, two values for the diameter of the 
bacteriophage have been calculated; one from the adsorption velocity constant 
and the other from the saturation capacity; the former yields an approximate 
particle diameter of 100 m/x, the latter a value of 60 to 120 m^u. A direct size 
determination of the same bacteriophage from its sedimentation velocity leads 
to a particle diameter of 80 to 90 myu- 



36 



THE GROWTH OF BACTERIOPHAGE 

By EMORY L. ELLIS and MAX DELBRtJCK* 

{From the William G. Kerckhof Laboratories of the Biological Sciences, California 
Institute of Technology, Pasadena) 

(Accepted for publication, September 7, 1938) 
INTRODUCTION 

Certain large protein molecules (viruses) possess the property of 
multiplying within living organisms. This process, which is at once 
so foreign to chemistry and so fundamental to biology, is exemplified 
in the multiplication of bacteriophage in the presence of susceptible 
bacteria. 

Bacteriophage offers a number of advantages for the study of the 
multiplication process not available with viruses which multiply at the 
expense of more complex hosts. It can be stored indefinitely in the 
absence of a host without deterioration. Its concentration can be 
determined with fair accuracy by several methods, and even the 
individual particles can be counted by d'Herelle's method. It can be 
concentrated, purified, and generally handled like nucleoprotein, to 
which class of substances it apparently belongs (Schlesinger (1) and 
Northrop (2)). The host organism is easy to culture and in some cases 
can be grown in purely synthetic media, thus the conditions of growth 
of the host and of the phage can be controlled and varied in a quanti- 
tative and chemically well defined way. 

Before the main problem, which is elucidation of the multiplication 
process, can be studied, certain information regarding the behavior of 
phage is needed. Above all, the "natural history" of bacteriophage, 
i.e. its growth under a well defined set of cultural conditions, is as yet 
insufficiently known, the only extensive quantitative work being that 
of Krueger and Northrop (3) on an a.ni[-staphylococcus phage. The 
present work is a study of this problem, the growth of another phage 
{a.nti-Escherichia coli phage) under a standardized set of culture 
conditions. 

* Fellow of The Rockefeller Foundation. 



Reprinted by permission of the authors and The Rockefeller 

Institute from The Journal of General Physiology, 22 (3), 

365-384, January 20, 1939. 



37 



366 GROWTH OF BACTERIOPHAGE 



EXPERIMENTAL 



Bacteria Culture. — Our host organism was a strain of Escherichia coli, which 
was kindly provided by Dr. C. C. Lindegren. Difco nutrient broth (pH 6.6- 
6.8) and nutrient agar were selected as culture media. These media were selected 
for the present work because of the complications which arise when synthetic 
media are used. We thus avoided the difficulties arising from the need for 
accessory growth factors. 

Isolation, Culture, and Storage of Phage. — A bacteriophage active against this 
strain of coli was isolated in the usual way from fresh sewage filtrates. Its homo- 
geneity was assured by five successive single plaque isolations. The properties of 
this phage remained constant throughout the work. The average plaque size on 
1.5 per cent agar medium was 0.5 to 1.0 mm. 

Phage was prepared by adding to 25 cc. of broth, 0.1 cc. of a 20 hour culture 
of bacteria, and 0.1 cc. of a previous phage preparation. After 3^ hours at 37° 
the culture had become clear, and contained about 10^ phage particles. 

Such lysates even though stored in the ice box, decreased in phage concentra- 
tion to about 20 per cent of their initial value in 1 day, and to about 2 per cent 
in a week, after which they remained constant. Part of this lost phage activity 
was found to be present in a small quantity of a precipitate which had sedimented 
during this storage period. 

Therefore, lysates were always filtered through Jena sintered glass filters (5 on 
3 grade) immediately after preparation. The phage concentration of these fil- 
trates also decreased on storage, though more slowly, falling to 20 per cent in a 
week. However, 1:100 dilutions in distilled water of the fresh filtered lysates 
retained a constant assay value for several months, and these diluted preparations 
were used in the work reported here, e.xcept where otherwise specified. 

This inactivation of our undiluted filtered phage suspensions on standing is 
probably a result of a combination of phage and specific phage inhibiting sub- 
stances from the bacteria, as suggested by Burnett (4, 5). To test this hypothesis 
we prepared a polysaccharide fraction from agar cultures of these bacteria, ac- 
cording to a method reported by Heidelberger et al. (6). Aqueous solutions of 
this material, when mixed with phage suspensions, rapidly inactivated the phage. 

Method of Assay. — We have used a modification of the plaque counting method 
of d'Herelle (7) throughout this work for the determination of phage concentra- 
tions. Although the plaque counting method has been reported unsatisfactory 
by various investigators, under our conditions it has proven to be entirely satis- 
factory. 

Phage preparations suitably diluted in 18 hour broth cultures of bacteria to 
give a readily countable number of plaques (100 to 1000) were spread with a bent 
glass capillary over the surface of nutrient agar plates which had been dried by 
inverting on sterile filter paper overnight. The plates were then incubated 6 to 
24 hours at 37°C. at which time the plaques were readily distinguishable. The 
0.1 cc. used for spreading was completely soaked into the agar thus prepared in 



38 



EMORY L. ELLIS AND MAX DELBRUCK 367 

2 to 3 minutes, thus giving no opportunity for the multiplication of phage in the 
liquid phase. Each step of each dilution was done with fresh sterile glassware. 
Tests of the amount of phage adhering to the glass spreaders showed that this 
quantity is negligible. 

The time of contact between phage and bacteria in the final dilution before 
plating has no measurable influence on the plaque count, up to 5 minutes at 25°C. 
Even if phage alone is spread on the plate and allowed to soak in for 10 minutes, 
before seeding the plate with bacteria, only a small decrease in plaque count is 
apparent (about 20 per cent). This decrease we attribute to failure of some 
phage particles to come into contact with bacteria. 

Under parallel conditions, the reproducibility of an assay is limited by the 
sampling error, which in this case is equal to the square root of the number of 
plaques (10 per cent for counts of 100; 3.2 per cent for counts of 1000). To test 
the effect of phage concentration on the number of plaques obtained, successive 
dilutions of a phage preparation were all plated, and the number of plaques 
enumerated. Over a 100-fold range of dilution, the plaque count was in linear 
proportion to the phage concentration. (See Fig. 1.) 

Dreyer and Campbell-Renton (8) using a different anti-co/i phage and an anti- 
staphylococcus phage, and a different technique found a complicated dependence 
of plaque count on dilution. Such a finding is incompatible with the concept 
that phage particles behave as single particles, i.e. without interaction, with 
respect to plaque formation. Our experiments showed no evidence of such a 
complicated behavior, and we ascribe it therefore to some secondary cause inherent 
in their procedure. 

Bronfenbrenner and Korb (9) using a phage active against B. dysenteriae Shiga, 
and a different plating technique found that when the agar concentration was 
changed from 1 per cent to 2.5 per cent, the number of plaques was reduced to 
1 per cent of its former value. They ascribed this to a change in the water 
supplied to the bacteria. With the technique which we have employed, variation 
of the agar concentration from 0.75 per cent to 3.0 per cent, had Httle influence 
on the number of plaques produced, though the size decreased noticeably with 
increasing agar concentration. (See Table L) 

Changes in the concentration of bacteria spread with the phage on the agar 
plates had no important influence on the number of plaques obtained. (See 
Table I.) The temperature at which plates were incubated had no significant 
effect on the number of plaques produced. (See Table L) 

In appraising the accuracy of this method, several points must be borne in 
mind. With our phage, our experiments confirm in the main the picture proposed 
by d'Herelle, according to which a phage particle grows in the following way: 
it becomes attached to a susceptible bacterium, multiplies upon or within it up 
to a critical time, when the newly formed phage particles are dispersed into the 
solution. 

In the plaque counting method a single phage particle and an infected bac- 
terium containing any number of phage particles will each give only one plaque. 



39 



368 



GROWTH OF BACTERIOPHAGE 



This method therefore, does not give the number of phage particles but the 
number of loci within the solution at which one or more phage particles exist. 
These loci will hereafter be called "infective centers." The linear relationship 
between phage concentration and plaque count (Fig. 1) does not prove that the 
number of plaques is equal to the number of infective centers, but only that it is 
proportional to this number. We shall call the fraction of infective centers which 



1280 














/ 


640 












/ 




^?0 










/ 


/ 




IRO 








A 


/ 






80 






/ 


/ 








40 




/ 


/ 










20 


A 


/ 












/ 


/ 















2 4 6 18 32 64 128 

Eelative concentration of phage 



256 



Fig. 1. Proportionality of the phage concentration to the plaque count. 
Successive twofold dilutions of a phage preparation were plated in duplicate 
on nutrient agar; 0.1 cc. on each plate. The plaque counts from two such series 
of dilutions are plotted against the relative phage concentration, both on a loga- 
rithmic scale. 

produces plaques the "efficiency of plating." With the concentrations of phage 
and bacteria which we have used this coefficient is essentially the fraction of 
infected bacteria in the suspension spread on the plate, which goes through to lysis 
under our cultural conditions on the agar medium. After plating, the phage 
particles released by this lysis infect the surrounding bacteria, increasing only the 
size, and not the number of plaques. 



40 



EMORY L. ELLIS AND MAX DELBRUCK 



369 



TABLE I 

Independence of Plaque Count on Plating Method 



Agar concentration 



Plates were prepared in which the agar strength varied, and all spread with 0.1 cc. 
of the same dilution of a phage preparation. There is no significant difference in the 
numbers of plaques obtained. 



Agar concentration, per cent. 
Plaque counts 



Average 

Plaque size, mm . 



0.75 


1.5 


3.0 


394 


373 


424 


408 


430 


427 


376 


443 


455 


411 


465 


416 


373 


404 


469 


392 


423 


438 


2 


0.5 


0.2 



Concentration of plating coli 



A broth suspension of bacteria (10' bacteria/cc.) was prepared from a 24 hour agar 
slant and used at various dilutions, as the plating suspension for a single phage dilution. 
There are no significant differences in the plaque counts except at the highest dilution of 
the bacterial suspension, where the count is about 15 per cent lower. 



Concentration 


Plaque count 


1 


920 


1/5 


961 


1/25 


854 


1/125 


773 



Temperature of plate incubation 



Twelve plates were spread with 0.1 cc. of the same suspension of phage and bacteria, 
divided into three groups, and incubated at different temperatures. There were no sig- 
nificant differences in the plaque counts obtained. 



Temperature, °C . 
Plaque count. . . . 



Average . 



37 
352 
343 
386 
422 

376 



24 
384 
405 
403 
479 

418 



10 

405 
377 
400 
406 

397 



41 



370 GROWTH OF BACTERIOPHAGE 

The experimental determination of the efficiency of plating is described in a 
later section (see p. 379). The coefficient varies from 0.3 to 0.5. This means 
that three to live out of every ten infected bacteria produce plaques. The fact 
that the efficiency of plating is relatively insensitive to variations in the tempera- 
ture of plate incubation, density of plating coli, concentration of agar, etc. indi- 
cates that a definite fraction of the infected bacteria in the broth cultures do not 
readily go through to lysis when transferred to agar plates. For most experiments 
only the relative assay is significant; we have therefore, given the values derived 
directly from the plaque counts without taking into account the efficiency of 
plating, unless the contrary is stated. 

Growth Measurements 

The main features of the growth of this phage in broth cultures of 
the host are shown in Fig. 2. After a small initial increase (discussed 
below) the number of infective centers (individual phage particles, 
plus infected bacteria) in the suspension remains constant for a time, 
then rises sharply to a new value, after which it again remains constant. 
Later, a second sharp rise, not as clear-cut as the first, and finally a 
third rise occur. At this time visible lysis of the bacterial suspension 
takes place. A number of features of the growth process may be 
deduced from this and similar experiments, and this is the main con- 
cern of the present paper. 

The Initial Rise 

When a concentration of phage suitable for plating was added to a 
suspension of bacteria, and plated at once, a reproducible plaque count 
was obtained. If the suspension with added phage was allowed to 
stand 5 minutes at 37°C. (or 20 minutes at 25°C.) the number of 
plaques obtained on plating the suspension was found to be 1.6 times 
higher. This initial rise is not to be confused with the first "burst" 
which occurs later and increases the plaque count 70-fold. After the 
initial rise, the new value is readily duplicated and remains constant 
until the start of the first burst in the growth curve (30 minutes at 
37° and 60 minutes at 25°). 

This initial rise we attribute not to an increase in the number of 
infective centers, but to an increase in the probability of plaque forma- 
tion {i.e. an increase in the efficiency of plating) by infected bacteria 
in a progressed state; that is, bacteria in which the phage particle has 
commenced to multiply. That this rise results from a change in the 



42 



EMORY L. ELLIS AND MAX DELBRUCK 



371 



efficiency of plating and not from a quick increase in the number of 
infective centers is evident from the following experiment. Bacteria 
were grown for 24 hours at 25°C. on agar slants, then suspended in 
broth. Phage was added to this suspension and to a suspension of 
bacteria grown in the usual way, and the concentration of infective 



a- 





































1^ 
















^ 


W 














i 

o 


o 














I 





lo* 








2-I5 
X 3-3 
+ 3-13 






+ ♦ ' 


o 

■f 







































Min.O 



20 



40 



60 



80 



100 

Time 



120 



MO 



150 



180 



200 



Fig. 2. Growth of phage in the presence of growing bacteria at 37°C. 
A diluted phage preparation was mixed with a suspension of bacteria con- 
taining 2 X 10* organisms per cc, and diluted after 3 minutes 1 to 50 in broth. 
At this time about 70 per cent of the phage had become attached to bacteria. 
The total number of infective centers was determined at intervals on samples of 
this growth mixture. Three such experiments, done on different days, are plotted 
in this figure. The same curve was easily reproducible with all phage prepara- 
tions stored under proper conditions. 

centers was determined on both. The initial value was 1.6 times 
higher in the agar grown bacteria than in the control experiment, and 
remained constant until actual growth occurred. The initial rise was 
therefore absent in this case, clearly a result of an increase in the 
efficiency of plating. A sufficient number of experiments were per- 



43 



372 GROWTH OF BACTERIOPHAGE 

formed with bacteria grown on agar to indicate that in other respects 
their behavior is similar to that of the bacteria grown in broth. The 
bacteria grown in this way on agar slants are in some way more sus- 
ceptible to lysis than the broth cultured bacteria. 

A dsorption 

The first step in the growth of bacteriophage is its attachment to 
susceptible bacteria. The rate of this attachment can be readily 
measured by centrifuging the bacteria out of a suspension containing 
phage, at various times, and determining the amount of phage which 
remains unattached in the supernatant {cf. Krueger (10)).^ 

According to the picture of phage growth outlined above, phage 
cannot multiply except when attached to bacteria; therefore, the rate 
of attachment may, under certain conditions, limit the rate of growth. 
We wished to determine the rate of this adsorption so that it could be 
taken into account in the interpretation of growth experiments, or 
eliminated if possible, as a factor influencing the growth rate. Our 
growth curves show that there is no increase in the number of infective 
centers up to a critical time; we could therefore, make measurements 
of the adsorption on living bacteria suspended in broth, so long as the 
time allowed for attachment was less than the time to the start of the 
first burst in the growth curve. The adsorption proved to be so rapid 
that this time interval was ample to obtain adsorption of all but a few 
per cent of the free phage if the bacteria concentration was above 
3 X 10^ The number of bacteria remained constant; the lag phase 
in their growth was longer than the experimental period. 

The rate of attachment was found to be first order with respect to 
the concentration of free phage (P/) and first order with respect to the 
concentration of bacteria {B) over a wide range of concentrations, in 
agreement with the results reported by Krueger (10). That is, the 
concentration of free phage followed the equation 

- '-^ = UPf)iB) 
at 

^ A very careful study of the adsorption of a co/z-phage has also been made 
by Schlesinger (Schlesinger, M., Z. Hyg. u. Infektionskrankh., 1932, 114, 136, 
149). Our results, which are less accurate and complete, agree qualitatively 
and quantitatively with the results of his detailed studies. 



44 



EMORY L. ELLIS AND MAX DELBRUCK 373 

in which ^a was found to be 1.2 X lO-^cm.Vmin. at 15° and 1.9 X 10 "^ 
cm.Vmin. at 25°C. These rate constants are about five times greater 
than those reported by Krueger (10). With our ordinary 18 hour 
bacteria cultures (containing 2 X 10^ B. coli/cc.) we thus obtain 70 
per cent attachment of phage in 3 minutes and 98 per cent in 10 
minutes. The adsorption follows the equation accurately until more 
than 90 per cent attachment has been accomplished, and then slows 
down somewhat, indicating either that not all the phage particles 
have the same afiinity for the bacteria, or that equilibrium is being 
approached. Other experiments not recorded here suggest that, if an 
equilibrium exists, it lies too far in favor of adsorption to be readily 
detected. This equation expresses the rate of adsorption even when a 
tenfold excess of phage over bacteria is present, indicating that a 
single bacterium can accommodate a large number of phage particles 
on its surface, as found by several previous workers (5, 10). 

Krueger (10) found a true equilibrium between free and adsorbed 
phage. The absence of a detectable desorption in our case may result 
from the fixation of adsorbed phage by growth processes, since our 
conditions permitted growth, whereas Krueger's experiments were 
conducted at a temperature at which the phage could not grow. 

Growth of Phage 

Following adsorption of the phage particle on a susceptible bac- 
terium, multiplication occurs, though this is not apparent as an in- 
crease in the number of plaques until the bacterium releases the 
resulting colony of phage particles into the solution. Because the 
adsorption under proper conditions is so rapid and complete (as shown 
above) experiments could be devised in which only the influence of the 
processes following adsorption could be observed. 

The details of these experiments were as follows: 0.1 cc. of a phage 
suspension of appropriate concentration was added to 0.9 cc. of an 18 
hour broth bacterial culture, containing about 2 X 10^ B. coli / cc. 
After standing for a few minutes, 70 to 90 per cent of the phage was 
attached to the bacteria. At this time, the mixture was diluted 50- 
fold in broth (previously adjusted to the required temperature) and 
incubated. Samples were removed at regular intervals, and the 
concentration of infective centers determined. 



45 



374 GROWTH OF BACTERIOPHAGE 

The results of three experiments at 37°C. are plotted in Fig. 2, and 
confirm the suggestion of d'Herelle that phage multiplies under a 
spatial constraint, i.e. within or upon the bacterium, and is suddenly 
liberated in a burst. It is seen that after the initial rise (discussed 
above) the count of infective centers remains constant up to 30 
minutes, and then rises about 70-fold above the initial value. The 
rise corresponds to the liberation of the phage particles which have 
multiplied in the initial constant period. This interpretation was 
verified by measurements of the free phage by centrifuging out the 
infected bacteria, and determining the number of phage particles in 
the supernatant liquid. The free phage concentration after adsorp- 
tion was, of course, small compared to the total and remained constant 
up to the time of the first rise. It then rose steeply and became sub- 
stantially equal to the total phage. 

The number of bacteria lysed in this first burst is too small a fraction 
of the total bacteria used in these experiments to be measured as a 
change in turbidity; the ratio of uninfected bacteria to the total 
possible number of infected bacteria before the first burst is 400 to 1, 
the largest number of bacteria which can disappear in the first burst is 
therefore only 0.25 per cent of the total. 

The phage particles liberated in the first burst are free to infect 
more bacteria. These phage particles then multiply within or on the 
newly infected bacteria; nevertheless, as before, the concentration of 
infective centers remains constant until these bacteria are lysed and 
release the phage which they contain into the medium. This gives 
the second burst which begins at about 70 minutes from the start of 
the experiment. Since the uninfected bacteria have been growing 
during this time, the bacteria lysed in the second burst amount to less 
than 5 per cent of the total bacteria present at this time. There is 
again therefore, no visible lysis. 

This process is repeated, leading to a third rise of smaller magnitude 
starting at 120 minutes. At this time, inspection of the culture, which 
has until now been growing more turbid with the growth of the unin- 
fected bacteria, shows a rapid lysis. The number of phage particles 
available at the end of the second rise was sufficient to infect the re- 
mainder of the bacteria. 

These results are typical of a large number of such experiments, at 



46 



EMORY L. ELLIS AND MAX DELBRUCK 375 

37°, all of which gave the 70-fold burst size, i.e. an average of 70 phage 
particles per infected bacterium, occurring quite accurately at the time 
shown, 30 minutes. Indeed, one of the most striking features of these 
experiments was the constancy of the time interval from adsorption 
to the start of the first burst. The magnitude of the rise (70-fold) 
was likewise readily reproducible by all phage preparations which 
had been stored under proper conditions to prevent deterioration (see 
above). 

Multiple Infection 

The adsorption measurements showed that a single bacterium can 
adsorb many phage particles. The subsequent growth of phage in 
these "multiple infected" bacteria might conceivably lead to (a) an 
increase in burst size; (b) a burst at an earlier time, or (c) the same 
burst size at the same time, as if only one of the adsorbed particles 
had been effective, and the others inactivated. In the presence of 
very great excesses of phage, Krueger and Northrop (3) and Northrop 
(2) report that visible lysis of the bacteria occurs in a very short time. 
It was possible therefore, that in our case, the latent period could be 
shortened by multiple infection. To determine this point, we have 
made several experiments of which the following is an example. 0.8 
cc. of a freshly prepared phage suspension containing 4 X 10^ particles 
per cc. (assay corrected for efficiency of plating) was added to 0.2 cc. 
of bacterial suspension containing 4 X 10^ bacteria per cc. The ratio 
of phage to bacteria in this mixture was 4 to 1. 5 minutes were 
allowed for adsorption, and then the mixture was diluted 1 to 12,500 
in broth, incubated at 25°, and the growth of the phage followed by 
plating at 20 minute intervals, with a control growth curve in which 
the phage to bacteria ratio was 1 to 10. No significant difference was 
found either in the latent period or in the size of the burst. The 
bacteria which had adsorbed several phage particles behaved as if 
only one of these particles was effective. 

Effect of Temperature on Latent Period and Burst Size 

A change in temperature might change either the latent period, i.e. 
the time of the burst, or change the size of the burst, or both. In 
order to obtain more accurate estimates of the burst size it is desirable 



47 



376 



GROWTH OF BACTERIOPHAGE 



to minimize reinfection during the period of observation. This is 
obtained by diluting the phage-bacteria mixture (after initial contact 
to secure adsorption) to such an extent that the rate of adsorption 
then becomes extremely small. In this way, a single "cycle" of growth, 
(infection, growth, burst) was obtained as the following example 




Min, 



Time 



Fig. 3. One-step growth curves. 
A suitable dilution of phage was mixed with a suspension of bacteria con- 
taining 2 X 10* organisms per cc. and allowed to stand at the indicated tempera- 
ture for 10 minutes to obtain more than 90 per cent adsorption of the phage. 
This mixture was then diluted 1 : 10^ in broth, and incubated. It was again diluted 
1 : 10 at the start of the first rise to further decrease the rate of adsorption of the 
phage set free in the first step. The time scales are in the ratio 1:2:6 for the 
temperatures 37, 25, and 16.6°C. Log P/Pq is plotted, Po being the initial con- 
centration of infective centers and P the concentration at time t. The broken 
line indicates the growth curve of the bacteria under the corresponding conditions. 



48 



EMORY L. ELLIS AND MAX DELBRUCK 377 

shows. 0.1 cc. of phage of appropriate and known concentration was 
added to 0.9 cc. of an 18 hour culture and allowed to stand in this con- 
centrated bacterial suspension for 10 minutes at the temperature of 
the experiment. This mixture was then diluted 1 : 10* in broth and 
incubated at the temperature chosen. Samples of this diluted mix- 
ture were withdrawn at regular intervals and assayed. The results 
of three such experiments are plotted in Fig. 3. The rise corresponds 
to the average number of phage produced per burst, and its value can 
be appraised better in these experiments than in the complete growth 
curve previously given (Fig. 2) where there is probably some over- 
lapping of the steps. In these experiments the rise is seen to be 
practically identical at the three temperatures, and equals about sixty 
particles per infected bacterium, but the time at which the rise oc- 
curred was 30 minutes at 37°, 60 minutes at 25°, and 180 minutes at 
16.6°. This shows that the effect of temperature is solely on the 
latent period. 

We have also made separate measurements of the rate of bacterial 
growth under the conditions of these experiments. They show that 
the average division period of the bacteria in their logarithmic growth 
phase varies in the same way with temperature, as the length of the 
latent period of phage growth. The figures are: 



Temperature 


Division period of B 


Latent period of P growth 


"C 
16.6 
25 
37 


min. 

About 120 
42 
21 


min. 
180 

60 

30 



There is a constant ratio (3/2) between the latent period of phage 
growth and the division period of the bacteria. This coincidence 
suggests a connection between the time required for division of a 
bacterium under optimum growth conditions, and the time from its 
infection by phage to its lysis. 

Individual Phage Particle 

The growth curves described above give averages only of large 
numbers of bursts. They can, however, also be studied individually, 
as was first done by Burnett (11). 



49 



378 



GROWTH OF BACTERIOPHAGE 



If from a mixture containing many particles very small samples are 
withdrawn, containing each on the average only about one or less 
particles, then the fraction pr of samples containing r particles is given 
by Poissons' (12) formula, 



Pr^ 



nTe' 



(1) 



where n is the average number of particles in a sample and e is the 
Napierian logarithm base. If the average number n is unknown, it 
can be evaluated from an experimental determination of any single 

TABLE II 

Distribution of Individual Particles among Small Samples 
A suitably diluted phage preparation was added to 5 cc. of 18 hour bacteria 
culture and 0.1 cc. samples of this mixture were plated. The distribution of 
particles among the samples is that predicted by formula (1). 





pT (experimental) 


Pr (calculated) 


plaques on 13 plates 

1 plaque " 14 " 

2 plaques " 5 " 

3 " "1 plate 

4 " " Opiates 


0.394 
0.424 
0.151 
0.033 
0.000 


0.441 
0.363 
0.148 
0.040 
0.008 


27 " " 33 " 


1.002 


1.000 



one of the pr, for instance from a determination of Po, the fraction of 
samples containing no particles: 

n = —Inpo (2) 

Let us now consider the following experiment. A small number of 
phage particles is added to a suspension containing bacteria in high 
concentration. Within a few minutes each phage particle has at- 
tached itself to a bacterium. The mixture is then diluted with a large 
volume of broth, in order to have the bacteria in low concentration so 
that after the first burst a long time elapses before reinfection, as in 
the one step growth curves. Samples (0.05 cc.) are removed from 
this mixture to separate small vials and incubated at the desired 
temperature. If these samples are plated separately (after adding a 
drop of bacterial suspension to each vial) before the occurrence of 
bursts, the fraction of the plates containing 0, 1,2, etc. plaques is found 



50 



EMORY L. ELLIS AND MAX DELBRUCK 379 

to conform to formula (1) (see Table II). In this experiment we 
could also have inferred the average number of particles per sample, 
using formula (2), from the fraction of the plates showing no plaques 
(giving 0.93 per sample) instead of from the total number of plaques 
(27/33 = 0.82 per sample). 

Experimental Measure of Efficiency of Plating 

If the samples are incubated until the bursts have occurred, and 
then plated, the samples which had no particles will still show no 
plaques, those with one or more particles will show a large number, 
depending on the size of the burst, and on the efSciency of plating. In 
any case, if we wait until all bursts have occurred, only those samples 
which really contained no particle will show no plaques, quite inde- 
pendent of any inefhciency of plating. From this fraction of plates 
showing no plaques we can therefore evaluate the true number of 
particles originally present in the solution, and by comparison with 
the regular assay evaluate the efficiency of plating. In this way we 
have determined our efhciency of plating to be about 0.4. For 
instance, one such experiment gave no plaques on 23 out of 40 plates, 
and many plaques on each of the remaining plates. This gives 

23 
Pa = —ov 0.57 from which n = 0.56 particles per sample. A parallel 

assay of the stock phage used indicated 0.22 particles per sample; 

0.22 
the plating efficiency was therefore tttz = 0.39. This plating effi- 
ciency remains fairly constant under our standard conditions for 
assay. The increase in probability of plaque formation which we 
suppose to take place following the infection of a bacterium by the 
phage particle, i.e. the initial rise, brings the plating efficiency up to 
0.65. 

The Burst Size 

Single particle experiments such as that described above, revealed a 
great fluctuation in the magnitude of individual bursts, far larger than 
one would expect from the differences in size of the individual bacteria 
in a culture; indeed, they vary from a few particles to two hundred 
or more. Data from one such experiment are given in Table III. 



51 



380 



GROWTH OF BACTERIOPHAGE 



We at first suspected that the fluctuation in burst size was con- 
nected with the time of the burst, in that early bursts were small and 
late bursts big, and the fluctuation was due to the experimental 
superposition of these. However, measurement of a large number of 
bursts, plated at a time when only a small fraction of the bursts had 
occurred, showed the same large fluctuation. We then suspected that 
the particles of a burst were not liberated simultaneously, but over an 
interval of time. In this case one might expect a greater homogeneity 



TABLE III 

Fluctuation in Individual Burst Size 
97.9 per cent of phage attached to bacteria in presence of excess bacteria (10 
minutes), this mixture diluted, and samples incubated 200 minutes, then entire 
sample plated with added bacteria. 



25 plates show plaques 
1 plate shows 1 plaque 
14 plates show bursts 
Average burst size, taking account of probable doubles = 48 


Bursts 


130 

58 

26 

123 

83 

9 

31 

5 

S3 
48 
72 
45 
190 
9 


Total 


882 plaques 





in burst size, if measurements were made at a late time when they are 
at their maximum value. This view also was found by experiment 
to be false. 

The cause of the great fluctuation in burst size is therefore still 
obscure. 

DISCUSSION 

The results presented above show that the growth of this strain of 
phage is not uniform, but in bursts. These bursts though of constant 



52 



EMORY L. ELLIS AND MAX DELBRUCK 381 

average size, under our conditions, vary widely in individual size. A 
burst occurs after a definite latent period following the adsorption of 
the phage on susceptible bacteria, and visible lysis coincides only 
with the last step-wise rise in the growth curve when the phage parti- 
cles outnumber the bacteria present. It seemed reasonable to us to 
assume that the burst is identical with the lysis of the individual 
bacterium. 

Krueger and Northrop (3), in their careful quantitative studies 
of an ant'i-staphylococcus phage came to an interpretation of their 
results which differs in some important respects from the above: 

1. Their growth curves were smooth and gave no indication of steps; 
they concluded therefore that the production of phage is a continuous 
process. 

2. In their case, the free phage during the logarithmic phase of a 
growth curve was an almost constant small fraction of the total phage. 
This led them to the view that there is an equilibrium between intra- 
cellular and extracellular phage. With an improved technique, 
Krueger (10) found that the fraction of free phage decreased in propor- 
tion to the growth of the bacteria, in conformity with the assumption 
of an equilibrium between two phases. 

3. Krueger and Northrop (3) found that visible lysis occurred when 
a critical ratio of total phage to bacteria had been attained, and they 
assumed that there was no lysis in the earlier period of phage growth. 

To appreciate the nature of these differences it must be born in 
mind that their method of assay was essentially different from ours. 
They used, as a measure of the "activity" of the sample of phage 
assayed, the time required for it to lyse a test suspension of bacteria 
under standard conditions. This time interval, according to the 
picture of the growth process given here, is the composite effect of a 
number of factors: the average time required for adsorption of free 
phage, its rate of growth in the infected bacteria, the time and size of 
burst, and the average time required for repetition of this process 
until the number of phage particles exceeds the number of bacteria 
and infects substantially all of them. Then, after a time interval 
equal to the latent period, lysis occurs. 

This lysis assay method tends to measure the total number of 
phage particles rather than the number of infective centers as the 



53 



382 GROWTH OF BACTERIOPHAGE 

following considerations show. Let us take a sample of a growth 
mixture in which is suspended one infected bacterium containing 
fifty phage particles. If this sample is plated, it can show but a 
single plaque. However, if the sample is assayed by the lysis method, 
this single infective center soon sets free its fifty particles (or more, if 
multiplication is still proceeding) and the time required to attain 
lysis will approximate that for fifty free particles rather than that for a 
single particle. 

Since the burst does not lead to an increase in the number of phage 
particles, but only to their dispersion into the solution, the lysis method 
cannot give any steps in the concentration of the total phage in a growth 
curve. On the other hand one might have expected a step-wise in- 
crease in the concentration of free phage. However, the adsorption 
rate of the phage used by Krueger (10) is so slow that the infection 
of the bacteria is spread over a time longer than the presumed latent 
period, and therefore the bursts would be similarly spread in time, 
smoothing out any steps which might otherwise appear. Moreover, 
their measurements were made at 30 minute intervals, which even 
in our case would have been insufficient to reveal the steps. 

The ratio between intracellular and extracellular phage would be 
determined, according to this picture of phage growth, by the ratio of 
the average time of adsorption to the average latent period. The 
average time of adsorption would decrease as the bacteria increased, 
shifting the ratio of intracellular to extracellular phage in precisely 
the manner described by Krueger (10). 

As we have indicated in the description of our growth curves, lysis 
of bacteria should become visible only at a late time. Infection of a 
large fraction of the bacteria is possible only after the free phage has 
attained a value comparable to the number of bacteria, and visible 
lysis should then set in after the lapse of a latent period. At this time 
the total phage (by activity assay) will be already large compared to 
the number of bacteria, in agreement with Krueger and Northrop's 
findings. 

It appears therefore that while Krueger and Northrop's picture 
does not apply to our phage and bacteria, their results do not exclude 
for their phage the picture which we have adopted. It would be of 
fundamental importance if two phages behave in such a markedly 
dilEferent way. 



54 



EMORY L. ELLIS AND MAX DELBRUCK 383 

SUMMARY 

1. An Sinti-Escherichia coli phage has been isolated and its behavior 
studied. 

2. A plaque counting method for this phage is described, and shown 
to give a number of plaques which is proportional to the phage concen- 
tration. The number of plaques is shown to be independent of agar 
concentration, temperature of plate incubation, and concentration of 
the suspension of plating bacteria. 

3. The efliciency of plating, i.e. the probability of plaque formation 
by a phage particle, depends somewhat on the culture of bacteria 
used for plating, and averages around 0.4. 

4. Methods are described to avoid the inactivation of phage by 
substances in the fresh lysates. 

5. The growth of phage can be divided into three periods: adsorp- 
tion of the phage on the bacterium, growth upon or within the bac- 
terium (latent period), and the release of the phage (burst). 

6. The rate of adsorption of phage was found to be proportional 
to the concentration of phage and to the concentration of bacteria. 
The rate constant ka is 1.2 X IQ-^ cm.Vmin. at 15°C. and 1.9 X IQ-' 
cm.Vniin. at 25°. 

7. The average latent period varies with the temperature in the 
same way as the division period of the bacteria. 

8. The latent period before a burst of individual infected bacteria 
varies under constant conditions between a minimal value and about 
twice this value. 

9. The average latent period and the average burst size are neither 
increased nor decreased by a fourfold infection of the bacteria with 
phage. 

10. The average burst size is independent of the temperature, and 
is about 60 phage particles per bacterium. 

11. The individual bursts vary in size from a few particles to about 
200. The same variability is found when the early bursts are mea- 
sured separately, and when all the bursts are measured at a late time. 

One of us (E. L. E.) wishes to acknowledge a grant in aid from Mrs. 
Seeley W. Mudd. Acknowledgment is also made of the assistance 
of Mr. Dean Nichols during the preliminary phases of the work. 



55 



384 GROWTH OF BACTERIOPHAGE 

REFERENCES 

1. Schlesinger, M., Biochem. Z., Berlin, 1934, 273, 306. 

2. Northrop, J. H., /. Gen. Physiol., 1938, 21, 335. 

3. Krueger, A. P., and Northrop, J. H., J. Gen. Physiol., 1930, 14, 223. 

4. Burnet, F. M., Brit. J. Exp. Path., 1927, 8. 121. 

5. Burnet, F. M., Keogh, E. V., and Lush, D., Australian J. Exp. Biol, and Med. 

Sc, 1937, 15, suppl. to part 3, p. 227. 
G. Heidelberger, M., Kendall, F. E., and Scherp, H. W., ./. Exp. Med., 1936, 64, 
559. 

7. d'Herelle, F., The bacteriophage and its behavior, Baltimore, The Williams 

& Wilkins Co., 1926. 

8. Dreyer, C, and Campbell-Renton, M. L., /. Palh. and Bad., 1933, 36, 399. 

9. Bronfenbrenner, J. J., and Korb, C, Proc. Soc. Exp. Biol, and Med., 1923, 21, 

315. 

10. Krueger, A. P., /. Gen. Physiol., 1931, 14, 493. 

11. Burnet, F. M., Brit, J. Exp. Path., 1929, 10. 109. 

12. Poisson, S. D., Recherches sur la probability des jiigements en matiere criminelle 

et en matiere civile, precedees des regies generales du calcul des probabilites, 
Paris, 1837. 



56 



THE GROWTH OF BACTERIOPHAGE AND LYSIS OF THE HOST 

By M. DELBRUCK* 

{From the William G. KerckhoJ' Laboratories of the Biological Sciences, California Institute 

of Technology, Pasadena) 

(Received for publication, January 29, 1940) 

Introduction and Statement of Main Result 

Bacteriophage grows in the presence of living susceptible bacteria. In 
many but not all cases the growth of phage leads finally to a lysis of the 
bacterial cells, a phenomenon which in dense cultures manifests itself to 
the naked eye as a clearing of the bacterial culture. The exact nature of 
the connection between the growth of the phage and the dissolution of the 
cells has been a subject of controversy since the original discoveries of 
d'Herelle in 1917. 

D'Herelle believed that lysis is the process by which the phage, which 
has grown within the bacterium, is liberated from the cell and dispersed in 
solution. Many later authors, notably Burnet, have concurred with him 
on this point. Last year Ellis and Delbriick (1) published detailed evi- 
dence that phage liberation in B. coli occurs in sudden bursts and showed 
that all the evidence was compatible with the assumption that in sensitive 
strains the bursts of phage liberation occurred only if and when a cell is 
lysed. 

Northrop and Krueger (3-5) on the other hand have developed ideas 
along a somewhat different line in the course of their extensive research 
with a strain of Staphylococcus aureus and a bacteriophage active against 
it. Bordet (2) had put forward the conception that phage production 
followed by lysis is a more or less normal physiological function of the 
bacteria. In lysogenic strains where visible lysis never occurs it can be 
put into close analogy with the production of an extra-cellular enzyme. 
Northrop's and Krueger's work served to substantiate this view also in 
their case where the phage growth leads finally to the dissolution of the 
bacteria. In their view lysis of the bacteria is a secondary and incidental 
activity of the phage. 

Krueger and Northrop (3) found first that clearing, if it occurs at all, 
begins when a certain threshold value in the ratio total phage/bacteria is 

* Fellow of The Rockefeller Foundation. 

Reprinted by permission of the author and The Rockefeller 

Institute from The Journal of General Physiology, 23 (5), 

643-660, May 20, 1940. 

57 



644 GROWTH OF PHAGE AND LYSIS 

overstepped. A considerable loss in total phage parallels the clearing of the 
culture. If sufficient phage is added so that the ratio phage/bacteria is 
greater than the threshold value lysis begins almost at once. This experi- 
ment was later repeated and confirmed by Northrop (4, 5) with purified 
concentrates of phage. 

Recently Northrop (5) found with a susceptible megatherium strain and 
homologous phage that the bulk of the phage was liberated before the cul- 
ture began to clear. He found further with a lysogenic strain which never 
showed clearing but produced phage lysing the sensitive strain, that the 
yield in phage from this lysogenic strain was large compared to the number 
of bacteria present in the culture. 

All these results indicate that in these strains lysis, if it occurs at all, is 
brought about by a mass attack of the phage on the bacteria after the phage 
have grown and been liberated into solution. 

We have now studied in more detail the relation between phage growth 
and lysis in a new sensitive strain of B. coli and homologous phage and have 
obtained results which may offer a basis for reconciling the two diverging 
lines of interpretation. 

We have found in this strain two entirely difTerent types of lysis, which 
we designate as "Lysis from within" and "Lysis from without." 

Lysis from without is brought about almost instantly by adsorption of 
phage at a threshold limit, which is equal to the adsorption capacity of that 
bacterium. No phage are liberated in this case, on the contrary, the ad- 
sorbed phage are lost. The phage attack the cell wall in such a way as to 
permit swelling of the cell, and its deformation into a spherical body. 

Lysis from within is brought about by adsorption of one (or few) phage 
particle(s). Under favorable conditions this one phage particle multiplies 
during a latent period within the bacterium up to a threshold value (which 
is equal to the adsorption capacity). When the threshold value is reached, 
and not before, the phage is liberated by a sudden destruction of the proto- 
plasmic membrane, which permits a rapid exudation of the cell contents 
without deformation of the cell wall. 

It would seem that the results of other observers may be explained by 
postulating that 

{a) In the case of Staphylococcus aureus the observable clearing is caused 
by lysis from without. Lysis from within either does not exist here and is 
replaced by continuous phage secretion ; or it exists but leads only to a slow 
equalization of the refractive indices of the cell interior and the milieu. 
The decrease of total phage during lysis is caused by the adsorption of 
phage in the process of lysis from without. 



58 



M. DELBRUCK 645 

(b) In the case of B. megatherium 36, sensitive, (Northrop (5)) the same 
relations hold. 

(c) In the case of B. megatherimn 899, lysogenic, (Northrop (5)), lysis 
from without does not occur, although the bacteria can adsorb a few phage 
particles each. Both the phage production capacity and the phage ad- 
sorption capacity are far smaller than the corresponding value for the same 
phage acting on the sensitive strain. 

The equation 

Adsorption capacity = maximum yield of phage per bacterium 

was found to hold true both for bacteria in the phase of rapid growth, and 
for saturated bacterial cultures that had been aerated for 24 hours and 
consisted only of very small bacteria. 

This equality points to a material connection between the bacterial con- 
stituents which can adsorb the phage and the new phage formed when it 
grows. These bacterial constituents we shall call b. It might be assumed 
that b, which the bacterium constantly produces without the help of phage 
(and in some cases also secretes), is part of the precursor which under 
favorable conditions is transformed into phage after combination of the 
bacterium with a phage particle from without. The complex bP might be 
the catalyst which in the cell transforms uncombined b into phage. The 
difference between a sensitive strain and a lysogenic strain would consist 
in this: in the sensitive strain the reaction 

catalyzed by bP 
b > phage (in the cell) 

would be faster than the production of b (in the cell). In the lysogenic 
strain b would be produced faster than it is converted into phage. This 
permits the bacterium and the phage to grow. 

The extremely interesting but puzzling observations of Burnet and McKie 
(6) on lysogenesis of different variants of one strain of B. enteriditis Gaertner, 
and of Burnet and Lush (7) on induction of resistance and lysogenesis by 
the phage in a strain of Staphylococcus albus may perhaps allow further 
analysis in the light of these speculations. 

EXPERIMENTAL 

The strains of B. coli and of homologous phage used in this work were obtained from 
the Pasadena Junior College, through the courtesy of Mr. F. Gardner. They have 
not been studied before and will be designated as B2 and P2, in distinction to the strains 
Bi and Pi used last year by Ellis and Delbriick (1). 

Growth curves of B2 in Difco nutrient broth at 37° (by colony counts) are shown in 
Fig. 1. The maximum division rate is considerably smaller than that of Bi. 



59 



646 



GROWTH OF PHAGE AND LYSIS 



10 r— 




o AERATED 
xNOT AERATED 



3 4 

TIME IN HOURS 

Fig. 1. Growth of bacteria in broth at 37°. 

The inoculating bacteria were taken from an 18 hour not aerated broth culture. 
Without aeration the growth reaches saturation at 10* B/cc. With aeration the growth 
proceeds further beyond 10^ B/cc. The maximum growth rate is in both cases equal 
and corresponds to an average division time of 30 minutes. The cessation of growth 
in the unaerated culture is therefore caused by lack of air. This is further supported 
by the top curve, which shows how the unaerated 18 hour culture, without the addition 
of fresh broth, proceeds to grow, when aerated, and reaches about 2 X 10^ B/cc. 



2 










•^ 


1 


o 
CQ 






J 


• 




o 




y* 


w 







p.-o 




>-•' 
5^-^ 


TIME IN 


HOURS 


^ 


2 


3 


4 


5 



Fig. 2. Growth of the standard experimental culture of bacteria in broth at 37°. 

At time zero 0.1 cc. of a 24 hour stock culture was added to 25 cc. broth. Every 
7.5 minutes platings were made for colony counts. The plotted values show complete 
agreement with Hershey's (11) finding that such aerated bacteria exhibit a sudden 
transition from the phase of cell enlargement (lag period) to the phase of cell division. 
The lag period is 2 hours, if defined as the time required for increase in cell number by a 
factor 1.5 (compare Hershey's discussion (11)). 



60 



M. DELBRUCK 647 

Synthetic Medium.— The bacteria and phage were also grown with aeration in a 
synthetic medium, consisting of 

/-Asparagine 2 gm. 

Glucose 4 gm. 

Na2HP04 (anhydrous) 6 gm. 

KH2PO4 3 gm. 

MgS04 .05 gm. 

NaCl 0.05 gm. 

Distilled H2O 1000 cc 

The bacteria in this medium grow more slowly than in broth but attain a higher final 
concentration. The phage also grow well on bacteria in this medium and cause lysis. 
But the growth rates of both the bacteria and the phage are only approximately re- 
producible with different batches of medium. These irregularities must be eliminated 
before the medium can be used for quantitative studies. 

Bacteria transferred from this medium into broth grow at once. Transferred from 
broth to this medium they require a period of about 24 hours of adaptation before 
growth begins. 

The bacteria were therefore carried on slants of synthetic medium agar and trans- 
ferred into broth only for the specific experiments. 

Stock phage was obtained by lysis in broth and filtration through Jena sintered 
glass filters. No measurable decrease in titer in periods over 6 months (in contrast to 
Pi, compare (1)). 

Phage assay by plaque count on Difco nutrient agar plates, as described previously 
(1). The plaques are large (2 mm. diameter) and are countable after 4 hours incuba- 
tion at 37°. 

Turbidity was determined by visual comparison with turbidity standards, in most 
cases taken from the same culture. Such turbidity determinations are of course very 
rough, but a refinement of technique in this respect did not seem profitable. The 
turbidity in any event would not be proportional to the lysis, which in dense cultures, 
as we have seen, is a complex phenomenon, in which changes in shape, size, and re- 
fractive index occur. Each of these factors contributes to the turbidity change. 

Preparation of Phage Concentrates.— Y or the experiments with large excess of 
phage over bacteria stock phage of very high titer were needed. These were obtained 
in the following way. 

It was observed that lysates obtained from synthetic medium cultures (in contrast to 
broth lysates) lost all their phage on filtration through Jena sintered glass filters. The 
phage is not inactivated by the filter but simply adsorbed, and it can be eluted with good 
yields by small volumes of distilled water. An example is given in Table I. 

These concentrates could be further concentrated by adding phosphate buffer to 
restore the original salt concentration and repeating the adsorption-elution procedure. 

Filters of the coarser grade 4 were also tried and though effective gave less reliable 
yields. 

Ground glass, silica, and fullers' earth were tried as adsorbents. These also gave 
good adsorption but the elution was again often unsatisfactory. 

The phosphate buffer was replaced by 1 per cent MgCl2 solution and by a 1 per cent 



61 



648 GROWTH OF PHAGE AND LYSIS 

NaCl solution. Both were as effective as phosphates in causing the phage to be ad- 
sorbed by the filter. 

Finally various concentrations of phosphate buffer were tried in order to deter- 
mine the lower limit at which adsorption would take place. It was found that reduction 
of the buffer concentration to one half practically eliminated the adsorption. 

By this method several concentrates were obtained with titers between 10"/cc. 
and 10^'^/cc. These concentrates were clear in transmitted light and showed intense 
blue Tyndall scattering. When kept in the ice box they showed no measurable decrease 
of titer over periods of more than 5 months. 

Standard Cullural Conditions. — The cultural conditions of the bacteria were stand- 
ardized in the following way. A stock culture was kept going for about a week by 
daily transfers into fresh synthetic medium. At the time the transfer was made, the 
culture contained about 2.5 X 10^ bacteria /cc. These 24 hour aerated bacteria were 
used for phage assay. They will be referred to as the stock culture. 

TABLE I 

Adsorption and Elution of Phage Jrom Jena Sintered Glass Filters 

Total phage 



25 CC. fresh lysate 1.4 X W> 35 X lOi" 

Filter lysate through Jena sintered glass filter grade 5 

on 3, filtrate 0.001 X lO" 0.025 X W 

Follow with 2 cc. distilled water 6.4 X W> 12 .8 X W 

Follow with 5 cc. distilled water 0. 19 X W 0.95 X 10" 

At the time of the transfer of the stock culture a sample of 0.1 cc. was also transferred 
into 25 cc. broth and aerated at 37°. This culture was used for growing the phage, 
and will be referred to as the experimental culture. 

Fig. 2 shows the growth of this experimental culture as determined by colony counts 
in 7.5 minute intervals. It is seen that the bacteria do not start to divide in the first 
90 minutes and divide at maximum rate after 3 hours. It can therefore be used to 
study phage growth either on bacteria that are small and have a long period of growth 
in size without divisions ahead of them, or on bacteria that have attained their maximum 
average size and divide at a maximum rate. 

Microscopic Observations 

We have made some microscopic studies of the lysis of our bacteria under 
various conditions, both in hanging drop preparations and on nutrient agar 
plates. The hanging drop preparations have the advantage that one can 
study changes in mobility of the bacteria and also that the bacteria are 
subjected to more uniform conditions, while on the agar plate one can 
follow the history of individual bacteria over a stretch of time. These 
observations were made at room temperature, the time schedule is therefore 
retarded in comparison with the growth curves at 37°C. 

Great differences in behavior were found depending on whether the bac- 



62 



M. DELBRUCK 649 

teria were infected with about an equal number of phage or with a large 
excess of phage (200 to 1 or still higher ratios) and these will be described 
separately. 

1. PIB = 1 

(a) Hanging Drop 

No changes in the shape of the bacteria were observed, only a gradual 
diminution in the number of bacteria. 

iJS) On Nutrient Agar Plates 

The bacteria were first mixed with phage in broth and aerated at 37° 
for 10 minutes, as in an ordinary growth curve set up. At 10 minutes a 
0.1 cc. sample was spread on an agar plate and observed under the micro- 
scope (magnification 600). A map was drawn of about 100 bacteria in the 
field of vision and these were checked every 5 minutes for changes. Up to 
about 30 minutes (now at room temperature, 25°) only a few bacteria dis- 
appeared, the others showed no change. Between 30 and 70 minutes 70 
per cent of the bacteria disappeared. They disappeared by a process of 
fading out, without noticeable change in shape. The fading out takes about 
2 minutes, a faint outline of the rod remaining visible for a long time after- 
wards. There appeared to be no correlation between the size of the bac- 
terium and the inception of its fading. Many bacteria which at the be- 
ginning had a constriction as if they were on the point of dividing behaved 
as one bacterium on lysis. Sometimes the fading appeared to start at one 
end of the rod and then to proceed gradually through its entire length. 

Discussion. — The bacteria are visible without staining under the micro- 
scope by virtue of a difference in the refractive index between their interior 
and the surrounding medium. The fading out without change of shape 
means then that the refractive index of its contents becomes equal to that 
of its milieu, while the cell wall retains its form. Inside and outside must 
suddenly become capable of free exchange. This must be due to a sudden 
disruption of the protoplasmic membrane. A change in permeability could 
hardly be so drastic as to permit complete equalization in so short a time. 

2. P/B = 200 or greater 
{a) Hanging Drop 

The bacteria kept their normal size and rod shape up to about 20 minutes. 
Then suddenly within 1 or 2 minutes the large majority was transformed 
into spherical bodies of about the same volume and small refractive power. 
These spherical bodies were visible for a long time and only gradually de- 
creased in number, some could be seen much distended and of oval shape. 

Besides these spherical bodies there appeared a few very minute rods that 
were extremely motile. 



63 



650 



GROWTH OF PHAGE AND LYSIS 



Also some of the spherical bodies showed great motility. 

(b) On Nutrient Agar Plates 

The bacteria were mixed with a 200-fold excess of phage in broth for 5 
minutes to permit time for adsorption. Then a 0.1 cc. sample was plated 
and observation began at 10 minutes. No changes in size or shape were 
observable up to 18 minutes. Then first few and soon many of the bacteria 
exhibited a variety of changes in form with parallel slow fading out. In 
most cases the rod simply swelled to an oval or spherical shape. Sometimes 
the swelling began at one end giving the impression of a rod attached to a 
little sphere, gradually the rod shortened and the sphere grew until only the 
sphere was left, which later assumed an irregular shape and finally faded out. 
Often the swelling began in the middle at the constriction of a dividing cell 

TABLE II 





No. of bacteria lysed by 


P/B 


Fading 


Swelling 


1.3 


83 









1 


66 









1 


59 









25 


43 






9 


150 


3 






15 


250 









12 


300 


1 






43 



and then extended to both ends until only one sphere was visible. The 
whole process from the inception of swelling to the attainment of a spherical 
shape took between 2 and 10 minutes. The spheres then fade out very 
slowly. 

Also a few minute rods are seen on the agar plate. They are never 
lysed. Their genesis has not been observed. 

Discussion. — It is clear from the above description that lysis under the 
influence of many external phage particles is an entirely different phe- 
nomenon from lysis under influence of one external phage particle which 
has grown within the bacterium to a large number. We must distinguish 
between lysis from within and lysis from without. In the latter case ap- 
parently the phage en masse attack the cell wall and so alter its elastic 
properties as to permit swelling of the cell and the uptake of water. Possi- 
bly the cell wall is actually dissolved and only the protoplasm remains and 
swells up freely. 



64 



M. DELBRUCK 651 

3. PIB between 1 and 200 

For these intermediate cases observations on agar plates showed not an 
intermediate type of lysis but a gradual shift of the fractions of bacteria 
that are lysed by fading or by swelling respectively. Table II illustrates 
this point. 

There is no ambiguity regarding the type of lysis a particular bacterium 
has undergone, except in rare cases when the fading has proceeded too far 
by the time of the next inspection, so that the form cannot be ascertained 
any more. 

Bronfenbrenner, Muckenfuss, and Hetler in 1927 (12) and Bayne-Jones and Sand- 
holzer in 1933 (13) have published very interesting photomicrographic moving pictures 
of lysing bacteria. They describe essentially the same morphological types of lysis 
which we find. In their experiments, however, the conditions of infection were not 
systematically varied and the ratio phage/bacteria was in no case determined. The 
significance of variations in the lytic process was therefore not recognized. 

One Step Growth Curves 

Phage growth curves with these strains of phage and bacteria show the 
same general features as those described by Ellis and Delbriick (1) for the 
strains Bi and Pi. If phage is added at time zero to an excess of bacteria 
the plaque count stays constant for 17 minutes, then rises, at first sharply 
and then more gently (on the logarithmic plot!) until about 30 minutes. 
At that time the first step is nearly completed. If the growth mixture has 
not been diluted before the beginning of this first rise another sharp rise 
begins at about 34 minutes (Fig. 3). If reinfection has been prevented by 
extreme dilution there is very little further increase in plaque count (see 
Fig. 2 of the preceding paper). 

It was decided to study in more detail this first step. The condition of 
extreme dilution under which the one step growth curve has to be measured 
facilitates an accurate analysis, because the samples which have to be 
plated at definite intervals need only be mixed with bacteria and plated, 
without further dilution. The time to which the assay is to be referred can 
then be defined within a fraction of a minute. 

Fig. 4 shows the values obtained from three such growth curves. It 
should be noted that since the plaque count is plotted directly (instead of 
logarithmically as in most plots of this kind) the sampling error (which is 
proportional to the measured value) is more conspicuous near the upper 
end of the growth curve. The actual percentage deviations are experi- 
mentally larger near the beginning of the rise, because here the plaque 



65 



652 



GROWTH OF PHAGE AND LYSIS 



count increases by a factor of twenty in 3 minutes and very slight inac- 
curacies in timing will entail huge percentage deviations in the plaque count. 
During the process of plating a bacterium may liberate the phage which it 
contains and thus add the full number of a "burst" to the normal sample 
value. This will place the point too high; it was probably the case in the 
17 minute point of the first growth curve. 



7' 



d 
o 








o 

X 


CO 
Ul 

n 






o 

X 


< 

Q. 




•P 

o 


^ 




3 


i. 








1 


< '0 


TIME 


IN Ml^ 


UTES 



10 



20 



30 40 



50 



Fig. 3. Phage growth on rapidly dividing bacteria. 

At time zero phage were added to the experimental culture of bacteria in its optimal 
growth phase (3 hours after inoculation) and well below its maximum concentration 
of bacteria (5 X 10^ against 2.5 X 10^). The initial phage concentration was 2 X 10"*, 
so that even after the first step the bacteria were still much in excess and no multiple 
infection occurred. After the second step the phage were in excess. The course of 
events resulting from this situation will be discussed separately. 

Dates of experiments: • 9-13, O 9-15, X 9-22. 

Fig. 4 brings out very clearly one point which was not recognizable in the 
logarithmic plots: Phage liberation starts suddenly after the latent period 
of 17 minutes and continues at a constant rate for about 16 minutes, at which 
point it ceases almost equally abruptly. In this interval from 17 to 33 
minutes the plaque count increases by a factor 170. 

These characteristics of phage growth, namely the latent period, the 
spread of the latent period, and the step size depend on the physiological 
state of the bacteria. For example, if, instead of using the above defined 
experimental culture after 3 hours when the bacteria are large and divide 
rapidly we had taken the bacteria from the stock culture directly, the one 
step phage growth curve under the same conditions (in broth at 37°C.) 



66 



M. DELBRUCK 



653 



203— 



o 
o 
\ 

Ul 

=5' 

a 
< 



IOC 



DATE 
A 9-28 
X 10-2 



D 10 -25 



10- 
|L-x- 



X 



A 

X A ' 

XD 



D a 

XX 



10 20 



TIM E 



N MINUTES 



30 



40 



50 



Fig. 4. One step growth curve of phage on rapidly dividing bacteria at 37°. Direct 

plot. 

At time zero about 10^ phage/cc. were added to 25 cc. of a rapidly growing broth 
culture of bacteria, that had been aerated for 3 hours and contained about 5 X 10^ 
bacteria/cc. After 5 minutes, 10^-, 10^-, and 10^-fold dilutions of this growth mixture 
were made in broth of 37° and these were further aerated and incubated. At 1 or 2 
minute intervals samples from these mixtures were plated for plaque counts. 

It is seen that the plaque count stays constant for 17 minutes, then increases linearly 
with time till 33 minutes when it reaches 170 times the original value. After 33 minutes 
it stays nearly constant. Phage liberation takes place uniformly during 16 minutes. 

It should be noted that on a logarithmic plot the rise would appear to be much more 
sudden. In fact on such a plot more than half of the step would be accomplished within 
3 minutes, when the plaque count has risen to twenty times the original value. 

TABLE III 

Characteristics of Phage Growth on Rapidly Dividing Bacteria and on 24 Hour Aerated 
Bacteria, Both Measured in Broth at 37° C, with Aeration 



Rapidly dividing bacteria . 
24 hrs. aerated bacteria . . 



Minimum 
latent period 



17 
30 



Spread of 
latent period 



16 

22 



Step size 



170 

20 



Saturation 
value 



250 
20 



would have been qualitatively the same but quantitatively quite different 
(see Fig. 5). Table III lists the respective values. 
The constant rate of phage Hberation in the best one step growth curves 



67 



654 



GROWTH OF PHAGE AND LYSIS 



30 



20- 



0* 



o 
o 

V) • 

UJ 


DATE- 
10-10 

IO-P4 






4 


> • 4 


_X - 


X 


y 


a 
< 

a. 






« 


• 
-X 


' X 

X 


> . 






t . 




- -X-X-' 


X'> 

• 
• 

• 
1 


TIME 


IN Mirv|UTES 







20 30 40 50 60 70 



80 



Fig. 5. One step phage growth on 24 hour aerated bacteria. Direct plot. 

At time zero 2 X 10^ phage/cc. were added to 1 cc. of a 24 hour stock culture of 
bacteria, containing 2 X 10^ bacteria/cc. After 5 minutes free phage were determined 
and suitable high dilutions in broth were further incubated at 37° with aeration. At 
intervals samples from these mixtures were plated for plaque count. 

It is seen that the plaque count stays constant for 30 minutes, then increases linearly 
with time till about 50 minutes when it reaches twenty times the original value. After 
50 minutes it stays nearly constant. Phage liberation takes place uniformly during 
20 minutes. 



g r"-A.-o s e 6 o 

ujd- ° 



3 
O 
< 

2 
o 
081- 



Lysfs 



,-->^' 



.TIME , IN (VHNUTE3 



5 10 15 20 25 30 35 40 

Fig. 6. Addition of a great excess of phage to a growing culture of bacteria, at 37*^ 





Date 


P/B 


log P/B 


P bound/5 


Free phage after 
adsorption 












per cent 


o 


9-13 


120 


2.08 


115 


4 


X 


9-22 


60 


1.78 


57 


5 


o 


10-26 


500 


2.70 


250 


50 


A 


11-20 


700 


2.85 


270 


60 



68 



M, DELBRUCK 655 

(Fig. 4) permits a closer analysis. Since the liberation of phage from the 
individual bacterium probably occurs quite suddenly when the bacterium is 
lysed our result means that the infected bacteria represent a mixture of 
groups with latent periods ranging between 17 and 33 minutes and that 
there is a uniform distribution of bacteria over this whole range of latent 
periods. 

The question arises as to what causes a bacterium to have a shorter or 
longer latent period. Several hypotheses might be suggested, either by 
ascribing the cause to statistical fluctuations of reactions involving a small 
number of particles (10), or by connecting it in one way or another with the 
bacterial cycle. The latter view seems to the author the more likely one 
but since it has not yet been worked out, further discussion will be deferred. 

Multiple Infection 

It was reported by Ellis and Delbriick (1) that if a bacterial suspension is 
infected with an excess of phage no changes occur in the latent period or in 
the burst size. At that time no phage concentrates were available and the 
maximum ratio of phage to bacteria attained in that work was only four 
to one. 

We have repeated this work with our new strains and with the concen- 
trates and have been able to work with much higher ratios of phage/bac- 
teria, up to 700 to 1. 

Fig. 6 shows some of the results obtained with the high ratios. Samples 
were assayed every 3 or 5 minutes. Since the assays here require several 
dilution steps these growth curves are less accurate both with respect to 
assay values and with respect to timing. 

The results show an initial decrease in plaque count because many phage 
particles are bound to one bacterium which then gives only one plaque. 
For instance, starting with 10^" phage/cc. and a hundred times less bacteria, 
one finds initially 10^° plaques/cc. After 10 minutes only 5 per cent of the 
phage will be left free; these will give 5 X 10^ plaques/cc. In addition the 
10^ bacteria/cc, each having adsorbed on the average 95 phage particles, 
will give 10^ plaques/cc, bringing the total plaques to 6 X lOVcc. If the 
initial ratio phage/bacteria is greater than a certain critical value the bac- 
teria show saturation. This saturation value depends on the physiological 
state of the bacteria. For instance, for rapidly growing bacteria, if the 
ratio is 500, the plaque count decreases only by a factor two. The satura- 
tion value is therefore 250. On the other hand, for 24 hour aerated bac- 
teria, the saturation value is only about 20. (See Table III, last column.) 

We have indicated in the figure the time during which clearing of the mix- 



69 



656 GROWTH OF PHAGE AND LYSIS 

ture occurs. If the initial ratio of phage/bacteria is smaller than the satu- 
ration value, clearing occurs only slightly earlier than in a one to one mixture, 
and it is accompanied by a noticeable increase in the plaque count. 

On the other hand if the initial ratio of phage/bacteria is greater than the 
saturation value, clearing occurs much earlier and is not accompanied by 
an increase in plaque count. 

In both cases the final plaque count is considerably smaller than the in- 
itial one; we have, in effect, a phage destruction by the adsorption that 
causes lysis. 

One can see the difference between the two types of lysis with the naked 
eye. A culture of rod shaped bacteria, like B. coli, shows flow lines on 
shaking due to the orientation of the rods under the influence of the shearing 
forces of unequal flow. In lyses under the influence of great excess of phage 
these flow lines disappear before the culture clears up, because the rods are 
transformed into spherical bodies before they disappear, as described in the 
section on microscopic observations. 

Growth of Phage and Lysis of Bacteria When Equivalent Numbers Are Mixed 

It can be predicted that a disturbance must arise when equivalent 
amounts of phage and bacteria are mixed, due to the fact that the phage 
that are liberated from the first lysing bacteria will cause an excess of phage 
over bacteria to be present. These phage will be adsorbed on bacteria that 
are already infected and will therefore not show up in a plaque count assay. 
They will moreover interfere with the phage growth in these bacteria and 
in some of them cause a lysis from without. 

Qualitatively the following can be predicted. We have seen that the 
phage will be liberated at a constant rate (after the lapse of the minimum 
latent period of 17 minutes). They will be adsorbed at a rate that is pro- 
portional to the phage concentration and to the bacterial concentration. 
The phage concentration is constantly increasing and the bacterial con- 
centration is constantly decreasing (due to lysis). The adsorption rate 
will therefore pass through a maximum and the net free phage production 
rate will pass through a minimum. The net result is the appearance of a 
point of inflection, i.e. a secondary step in the phage growth curve, in some 
cases even a temporary decrease in the free phage if the rate of adsorption 
at any time exceeds the rate of phage liberation. Because of the loss of 
phage by adsorption and partial lysis from without the total step size must 
be smaller than in a one step growth curve where the bacteria are in excess 
and where multiple adsorption is prevented by extreme dilution, after ad- 
sorption of the parent phage. 



70 



M. DELBRUCK 



657 



These predictions are borne out by the experimental results. Fig. 7 
shows three such growth curves where nearly equivalent amounts of phage 
and bacteria were mixed at time zero. The diminished yield is very 
pronounced and the secondary step is discernible in two sets of observational 
points. The condition of single infection of all bacteria at zero time can of 
course be realized only approximately. Even if exactly equivalent amounts 



200 
150 
100 

50 

10 



pTb" 

S X 1.4 
\ ^ 2.6 

CO 
Hi 

3 



cleoring 
I — "—I 



10 




20 TIME 30 IN MINUTES 50 



Fig. 7. Growth of phage if equivalent high concentrations (-^10 ^/cc.) of phage and 
bacteria are mi.xed at time zero. Direct plot. 

Besides the experimental points from three growth curves three theoretical curves 
are drawn in the figure. These are 

1. A one step growth curve with B in excess, taken from Fig. 4. 

2. A calculated growth curve, assuming inactivation of the liberated phage on 
bacteria not yet lysed. 

3. Same, but assuming that the adsorbing power of the bacterial constituents respon- 
sible for it is unimpaired till the completion of phage liberation and then vanishes 
abruptly. 

The time interval from the beginning of clearing to its completion is indicated. It 
falls well on the ascending part of the one step growth curve. In the one-to-one growth 
curves this ascending part is soon counter-balanced by the multiple adsorption loss, 
so that clearing seems to occur during a phase of little phage liberation. 

were mixed, the phage would not infect all the bacteria, but distribute 
themselves according to the probability formulas derived by Poisson. If 
P/B = n there will be a fraction e~" of the bacteria uninfected. On the 
other hand our phage assays, though fully reliable as far as relative values 
go, are not as certain with respect to absolute value, because of the difficulty 
of obtaining an accurate determination of the efficiency of plating {cf. 
Ellis and Delbruck (1)). 

It is not possible to make a complete quantitative prediction of the growth curve 
because it is not known in detail how the adsorption en masse of phage to a bacterium 



71 



658 GROWTH OF PHAGE AND LYSIS 

that is already near a lysis from within will interfere with this process. It is also not 
quite certain whether those parts of the surface of the bacterium that adsorb the phage 
will lose their capacity of binding phage immediately upon lysis. In the strains used 
previously a slow decrease of phage assay after lysis could be ascribed to the continued 
"adsorption" of phage onto those scattered surface elements. No such decrease of 
phage assay was ever observed with the new strain. But such observations refer only 
to inactivation long after lysis and do not tell us whether the adsorbent is instantly 
destroyed upon lysis. 

We have therefore calculated growth curves on the basis of two extreme assumptions. 

(a) The amount of adsorbent decreases linearly from its initial value to zero during 
the 16 minutes in which the bacteria are lysed. 

(b) The amount of adsorbent stays constant at its initial value throughout the course 
of lysis. 

Case (c) is described by the differential equation 

dP/dt = A - kBo (i-t/T) P 
In case (b) we have 

dP/dt = A - kBaP 

In these equations the first term, A, represents the phage liberation by lysis during 
the interval T, as determined in the one step growth curves, the second term is the 
decrease of phage due to adsorption either on the unlysed bacteria only (case a) or on 
the unlysed bacteria plus the adsorbent from the lysed bacteria (case b). 

These equations can be integrated explicitly. 

We obtain in case (a) 

P = M\/TTe2(r/-)'-i^-')'/-'[G(r/r) - G{[T - t]/r)] 
with 

T = \/2T/kBo 
and G(x) the Gaussian integral 

G(x) = 4- f e-'' dx 
VT Jo 

In case (b), with constant adsorbent, the adsorption rate grows continuously with 
the free phage concentration. In this case we have therefore no point of inflection 
but a continuous asymptotic approach to the final titre 

kBo 

Since all required constants are known from independent experiments, the par- 
ticular solutions applying to our case can be evaluated quantitatively. These have 
been plotted in Fig. 6. The experimental values fall between the limits set by these 
two cases. 

We have also plotted the curve obtained in an ordinary one step growth 
curve, with B in excess (taken from Fig. 4). The difference between this 



72 



M. DELBRUCK 



659 



curve and the experimental values is the amount of phage lost by adsorption. 
The data show that about one hundred phage are lost per bacterium. It is 
clear that this loss depends entirely on the rate of adsorption, which is 
determined by the product kBo. If we wish to increase the yield per bac- 
terium we have to decrease either Bo or k. Reduction of Bo brings us to 
the condition employed in the one step growth curves. Here the maxi- 
mum yield of phage growth per bacterium is obtained, but the actual con- 
centrations of phage are of course very small. 

A promising way of increasing the end titre of phage would be to reduce 
k, the adsorption rate constant. The very interesting experiments of 
Krueger and his coworkers (8, 9) on the influence of the addition of salts 
(NaCl and Na2S04) to a growth mixture of phage and bacteria would seem 
to be completely in accord with the assumption that the adsorption rate 
constant is diminished in the presence of salt. 

In fact, a diminished adsorption rate constant should manifest itself in 
several ways in a phage growth curve, in which one starts with low concen- 
trations of both phage and bacteria. Namely 

1. Delayed clearing, due to delayed adsorption, and therefore delayed 
phage growth. 

2. Higher maximum concentration of bacteria, due to delayed lysis. 

3. Higher end titre of phage, due to 

(a) higher number of bacteria producing phage 

(b) reduced loss of phage by multiple adsorption. 

4. Higher ratio of free to bound phage during the stationary growth 
phase, due to the fact, that every phage particle spends a longer time in the 
free state between liberation and adsorption. 

5. A period of constant bacterial concentration preceding lysis, when all 
bacteria are infected and cease to divide but when the phage concentration 
is not yet sufficient for lysis. 

Precisely these five differences from the normal course and no others were 
noted by Krueger and Strietmann (9) in their study of the influence of the 
addition of Na2S04. 

SUMMARY 

1. A new strain of B. coli and of phage active against it is described, and 
the relation between phage growth and lysis has been studied. It has been 
found that the phage can lyse these bacteria in two distinct ways, which 
have been designated lysis from within and lysis from without. 

2. Lysis from within is caused by infection of a bacterium by a single 
phage particle and multiplication of this particle up to a threshold value. 



73 



660 GROWTH OF PHAGE AND LYSIS 

The cell contents are then liberated into solution without deformation of 
the cell wall. 

3. Lysis from without is caused by adsorption of phage above a threshold 
value. The cell contents are liberated by a distension and destruction of 
the cell wall. The adsorbed phage is not retrieved upon lysis. No new 
phage is formed. 

4. The maximum yield of phage in a lysis from within is equal to the 
adsorption capacity. 

5. Liberation of phage from a culture in which the bacteria have been 
singly infected proceeds at a constant rate, after the lapse of a minimum 
latent period, until all the infected bacteria are lysed. 

6. If the bacteria are originally not highly in excess, this liberation is soon 
counterbalanced by multiple adsorption of the liberated phage to bacteria 
that are already infected. This leads to a reduction of the final yield. 

The author wishes to express his appreciation for the hospitality extended 
to him by the Biology Department of the California Institute of Technology 
during the tenure of a Fellowship of The Rockefeller Foundation. In 
particular he wishes to record his indebtedness to Dr. E. L. Ellis for constant 
help and advice and to Mr. F. Gardner for technical assistance. 

REFERENCES 

1. Ellis, E. L., and Delbruck, M., /. Gen. Physiol, 1939, 22, 365. 

2. Bordet, J., Proc. Roy. Soc. London, Series B, 1931, 107, 398. 

3. Krueger, A. P., and Northrop, J. H., /. Gen. Physiol, 1930, 14, 223. 

4. Northrop, J. H., /. Gen. Physiol, 1937, 21, 335. 

5. Northrop, J. H., /. Gen. Physiol, 1939, 23, 59. 

6. Burnet, F. M., and McKie, M., Australian J. Exp. Biol and Med. Sc, 1929, 6, 277. 

7. Burnet, F. M., and Lush, D., Australian J. Exp. Biol, and Med. Sc, 1936, 14, 27. 

8. Scribner, E. J., and Krueger, A. P., /. Gen. Physiol, 1937, 21, 1. 

9. Krueger, A. P., and Strietmann, W. L., /. Gen. Physiol, 1939, 22, 131. 

10. Delbruck, M., /. Physic. Chem., 1940, 8, 120. 

11. Hershey, A. D., /. Gen. Physiol, 1939, 23, 11. 

12. Bronfenbrenner, J., Muckenfuss, R. S., and Hetler, D. M., Am. J. Path., 1927, 3, 562. 

13. Bayne-Jones, S., and Sandholzer, L. A., /. Exp. Med., 1933, 57, 279. 



74 



THE INTRACELLULAR GROWTH OF BACTERIOPHAGES 

I. Liberation of Intracellular Bacteriophage T4 by Premature Lysis 
WITH Another Phage or with Cyanide 

By a. H. DOERMANN*. t 

{From the Department of Genetics, Carnegie Institution of Washington, 
Cold Spring Harbor) 

(Received for publication, August 20, 1951) 

Direct studies of bacteriophage reproduction have been handicapped by the 
fact that the cell wall of the infected bacterium presents a closed door to the 
investigator in the period between infection and lysis. As a result it was im- 
possible to demonstrate the presence of intracellular phage particles during this 
so called latent period, and, much less, to estimate their number or to describe 
them genetically. This barrier has now been penetrated. It is the purpose of the 
first two papers of this series to describe two methods for disrupting infected 
bacteria in such a way that the intracellular phage particles can be counted and 
their genetic constitution analyzed. 

The first method used to liberate intracellular bacteriophage depends on the 
induction of premature lysis in infected bacteria by "lysis from without" 
which occurs when a large excess of phage particles is adsorbed on bacteria (1). 
It was found by nephelometric tests that T6 lysates are efficient in disrupting 
cells when moderately high multiplicities are used (2). The further observation 
was made that the addition of a large number of T6 particles to bacteria previ- 
ously infected with T4, would, under some conditions, cause liberation of T4 
particles before the expiration of the normal latent period of these cells. It 
therefore seemed hopeful that a method of reproducibly disrupting infected 
bacteria could be developed on the basis of this preliminary knowledge. 

* The experiments described here were carried out while the author was a fellow 
of the Carnegie Institution of Washington. The author is indebted to Dr. M. Demerec 
and the staff of the Department of Genetics of the Carnegie Institution of Washington 
for providing facilities for this work. In particular the stimulating discussions with 
Dr. Barbara McClintock are gratefully acknowledged. The manuscript was prepared 
while the author held a fellowship in the Department of Biology of the California 
Institute of Technology. He is grateful to Dr. G. W. Beadle and the staff of that de- 
partment for their interest in this work, and especially to Dr. Max Delbriick for 
criticism of the manuscript. 

X Present address: Biology Division, Oak Ridge National Laboratory, Oak Ridge, 
Tennessee. 

Reprinted by permission of the author and The Rockefeller 

Institute from The Journal of General Physiology, 35 (4), 

645-656, March 20, 1952. 

75 



646 



INTRACELLULAR GROWTH OF BACTERI0PH.4GES. I 



The first experiments in devising a method of this kind were made with 
phage T5 (3). It was found that T5 is Hberated before the end of the latent 
period if the infected cells are exposed to a high excess of T6. However, the 
extremely low rate of adsorption of T5 coupled with difficulties in inactivation 
of unadsorbed phage by specific antisera indicated that this phage was a poor 
choice. Hence T4 was chosen because of its fast rate of adsorption and because 
of the availability of high titer antisera against it. The first experiments with 
T4, along with the T5 results, showed conclusively that, by itself, lysis from 
without is not sufficiently rapid for the purpose of this investigation. It is likely 
that phage growth continues after the addition of the lysing agent T6. There- 
fore the attempt was made to stop phage growth while T6 was allowed to ac- 
complish lysis from without. Low temperature could not be used for this purpose 



TABLE I 

Composition of the Growth Medium 



Material 



KH0PO4 

Na2HP04 (anhydrous) 

NH4CI 

MgS04-7HOH 

Glycerol 

Acid-hydrolyzed casein 

rf/- Tryptophan 

Gelatin* 

Tween-80 

* To reduce surface inactivation of free phage particles (4) 



Amount 



gm. per liter 
1.5 
3.0 
1.0 
0.2 
10.0 
5.0 
0.01 
0.02 
0.2 



since it also inhibits lysis from without. A search for a suitable metabolic in- 
hibitor was therefore undertaken, and cyanide was eventually chosen as the 
most suitable one. 

Materials and Methods 

The experiments described here were carried out with the system T4r48^ growing at 
37°C. in Escherichia coli, strain B/r/l. The latter is a Tl resistant, tr>ptophan-de- 
pendent mutant of B/r obtained from Dr. E. M. Witkin. 

Two media were used in these experiments, namely the growth medium and the 
lysing medium. The composition of the growth medium used for both bacterial and 
phage cultures, is given in Table I. The lysing medium consists of growth medium with 



^ The subscript refers to a particular r mutant of T4 which arose by mutation and 
was numbered after the system of Hershey and Rotman (12) using the high subscript 
number to avoid confusion with the mutants already described by Hershey and 
Rotman. 



76 



A. H. DOERMANN 647 

the addition of one part in ten of a high titer T6 phage filtrate (concentration of T6 
in lysing medium was ca. 4 X 10^ particles per ml.) and cyanide brought to a final 
concentration of 0.01 M. Specially designed experiments showed that at this concentra- 
tion the cyanide does not inactivate free phage particles, nor does the amount which 
reaches the plate affect titration by interference with plaque development. 

T6 was used as the lysing phage because in several experiments it proved to be a 
more effective lysing agent than any of the other T phages tested. Since only single 
stocks of the phages were compared in the early experiments, the superiority of T6 
over the other phages may have been due to a difference in the particular stock used, 
and not to an inherent difference among the phages. In fact, later experiments with 
different T6 stocks showed marked differences in lysing efficiency, and phage titer 
proved to be a poor criterion of lysing ability. The experiments described here were 
made with T6 stocks selected for their ability to induce lysis from without. The se- 
lections were made on the basis of nephelometric comparisons. 

Platings were made in agar layer (0.7 per cent agar) poured over nutrient agar plates 
(1.3 per cent agar), and in order to assay T4 in the presence of high titer T6, the in- 
dicator strain, B/6, was used. B/6 is completely resistant to T6 (no host range mutants 
have so far been found which will lyse the strain used here) and gives full efficiency 
of plating (compared to B) with T4. 

EXPERIMENTAL 

Experiments with the Stayidard Lysing Medium. — The experimental procedure used 
consisted essentially of a one-step growth experiment (5) with certain modifications. 
B/r/1 cells in the exponential growth phase were concentrated by centrifugation to 
about 10^ cells per ml. To these concentrated bacteria T4r48 was added and this ad- 
sorption mixture was incubated for 1 to 2 minutes with aeration, allowing at least 80 
per cent of the phage to be adsorbed to the bacteria. Then a 40-fold or larger dilution 
was made into growth medium containing anti-T4 rabbit serum. The serum inactivated 
most of the residual unadsorbed phage. After several minutes' incubation in the serum 
tube, a further dilution was made to reduce the serum concentration to one of relative 
inactivity. The resulting culture will be referred to as the source culture (SC). The 
entire experiment was carried out with the infected bacteria from SC. The titer of 
infected B/r/l in this tube was approximately 10^ cells per ml. 

Simultaneously with the dilution into the tube containing serum, another dilution 
from the adsorption tube was made. From the latter an estimation of the unadsorbed 
phage was made by assaying the supernatant after sedimentation of the cells. This 
step permits calculation of the multiplicity of infection (5). 

From SC a further dilution of 1:20 was made at some time before the end of the 
latent period. The resulting culture, containing approximately 5 X 10' infected bac- 
teria per ml., was used for determining the normal end of the latent period and for 
estimating the average yield of phage per infected cell. It will be called the control 
growth tube (GT). In addition, a number of precisely timed 20-fold dilutions were made 
from SC into lysing medium. These were titrated after they had been incubated in 
the lysing medium for 30 minutes or longer. Serial platings from the lysing medium 
cultures over a longer period of time have shown that the phage titer remained constant 
after 30 minutes' incubation. The titer calculated from these platings, divided by the 



77 



648 INTRACELLULAR GROWTH OF BACTERIOPHAGES. I 

titer of infected bacteria given by the preburst control platings, gives the average yield 
per infected cell. As a working hypothesis, this yield was considered to be the average 
number of intracellular phage particles per bacterium at the time of dilution into the 
lysing medium. Dividing these numbers by the control burst size gives the fraction of 
the control yield found in the experimental lysing medium tubes. 

The results of several typical experiments are shown in Fig. 1 in which the 
data are plotted on semilogarithmic coordinates. The fraction of the control 
yield found in a given experimental culture is plotted against the time at which 
the dilution is made into lysing medium. Curve 1 shows the results from a single 
experiment in which the bacteria were infected with an average of 7 phage 
particles each. Curve 2 is the composite result of four experiments in which the 
bacteria were infected with single phage particles. Curve 3 is the control one- 
step growth curve derived from the control growth tube platings in the four 
experiments of curve 2. 

Several striking results can be seen in these experiments. First, it is clearly 
seen that during the early stages of the latent period the virus-host complex is 
inactivated by the cyanide-T6 mixture, and that not even the infecting particles 
are recovered. Even when 7 phage particles were adsorbed on each bacterium, 
less than two are recovered per cell at the earliest stage tested, and the shape of 
the curve suggests that if earlier stages had been tested, still fewer would have 
been recovered. In experiments with singly infected bacteria, the earliest tests 
indicated that less than one infected bacterium in 80 liberated any phage at all. 
A second point to be noted is that the multiplicity of infection appears to 
influence slightly the time at which phage particles can be recovered from the 
cell, and it continues to affect the fraction found in the bacteria at a given time. 
That this difference is a real one seems clear from the consistency among the 
points of curve 2. This result has been observed in each experiment, although 
the effect appeared to be less pronounced in some experiments made at the 
lower temperature of 30°C. Attention should also be drawn to the fact that the 
shape of the curves is clearly not exponential. In fact, it parallels with a delay 
of several minutes the approximately linear DNA increase observed in this 
system (6). 

In connection with the preceding experiments a test was made to establish 
whether the cyanide concentration chosen was maximally effective in inhibiting 
phage synthesis. Using the described technique, but changing the cyanide con- 
centration of the lysing medium from 0.01 m to 0.004 m and 0.001 m in parallel 
aliquots, no difference was detectable in the three lysing media. Thus 0.01 m 
cyanide is well beyond the minimum concentration necessary and was con- 
sidered to be adequate for these experiments. 

Action of Cyanide in the Absence of the Lysing Agent T6. — Cohen and Anderson 
(7) reported a loss of infectious centers when infected bacteria were incubated 
in the presence of the antimetabolite 5-methyl tryptophan. Although the details 



78 



A. H, DOERMANN 



649 




79 



650 



INTRACELLULAR GROWTH OF BACTERIOPHAGES. I 



of their experiments differed somewhat from those presented here, tlie loss of 
infectious centers in their experiments suggested testing whether cyanide could 
cause a similar loss of infected bacteria in the present procedure. An experi- 
ment w-as made which was identical with the standard cyanide lysis experiment 



100- 



uj 10- 



« 1.0- 



T4r^AT 37' 



001- 




INFECTEO BACTERIA 

• -ON" 

O -CN~+ T€ 



I0~ 15 20 25 

MINUTES OF INCUBATION IN GROWTH MEDIUM AT 37* 



Fig. 2. The comparative effect of cyanide alone and cyanide plus T6 on singly 
infected bacteria at various stages in the latent period. 

except that T6 was omitted from one set and included in a parallel set of lysing 
medium cultures (Fig. 2). As in the case of 5-methyltryptophan, it is seen that 
cyanide alone caused a loss of infectious centers when added in the early stages 
of phage growth, although the loss is less than that produced by cyanide and 
T6 together. Furthermore, in the second half of the latent period, comparison 
of the two media showed clearly and surprisingly that a definite rise in titer of 
infective centers occurred even when the lysing agent, T6, was omitted from the 



80 



A. H. DOERMANN 651 

lysing medium. In fact, during the second half of the latent period, phage 
liberation is identical in the two media. 

In order to see whether lysis is actually occurring and can account for the 
liberation of phage, a nephelometric experiment was made introducing CN~ at 
two points in the latent period. Three cultures of B/r/1 growing exponentially 
in growth medium were infected with T4r48 (ca. fivefold multiplicity). One 
culture served as a control for normal lysis. To the second culture cyanide (0.01 
M final concentration) was added 7.5 minutes after addition of the virus and to 



CN~ADDEDT0 2 T4i. AT 37° 




10 20 30 40 50 60 70 80 90 290 

MINUTES AFTER ADDITION OF PHAGE TO BACTERIAL CULTURES 

Fig. 3. The turbidity of T4-infected bacterial cultures as affected by addition of 
cyanide at two stages in the latent period. 

the third tube 17.5 minutes after addition of the T4r48. The turbidities of these 
three cultures were followed with a nephelometer designed like that described 
previously by Underwood and Doermann (8), but with four separate units 
which permit independent readings on the four tubes without removing any of 
them from the instrument. The results indicate that CN~ added to infected 
bacteria early during the latent period does not induce lysis (Fig. 3). From the 
plaque count experiment (Fig. 2) it is seen that a loss of infective centers does 
occur. This loss must therefore be due to some cause other than lysis of these 
cells. In the later stages of the latent period, the turbidimetric experiment 
indicates that lysis occurs promptly upon the addition of CN~ to the culture 



81 



652 



INTRACELLULAR GROWTH OF BACTERIOPHAGES. I 




<u 


■>-> 


(U 


rfl 


o 


C 


-n 


■*-* 


C 


<U 


n1 


U 


ij 




fi 


B 


o 


to 




a, -o 


r 




<u 


03 






.t-i 


XI 




CJ 


c 


'^ 




4J 
J3 






H 




en 


C 

<u 




O 




C 







<u _c -^ 



3Aano HiMoao noaiNOo ni a"i3iA ivnm jo iN30 a3d 







'5) ^ 




>. 




'35 


^ 




^ 


o 


a; 


O 














-a 


^c 


be 






<u 


x: 


a 




1 


1 


o 


4J 
in 




4J 


m 


t> 


U 










c 




5J 


rt 


O 


O 




c« 


CN 


(X 

B 


~o 




a 


<U 


L« 






> 


o 


«3 






Ui 


tJ 


"O 




"2 


3 




3 




_o 


O 


OS 


o3 






T3 


tn 


ir. 




G, 


C 


<r) 


4J 






rt 


4J 

> 


-3 




-5 


C 








3 

u 


.22 






^ 




aj 




o 


c 


aJ 






-C 






3 




W) 


^ 


^ 


O 




3 

OJ 


OJ 


c« 




3 
TJ 


> 


o 


3 
4j 








tn 


£ 






o 




'E 

a; 




% 


'-3 


-i-i 
C 


a 








o 




3 


"3 


4) 






a 

o 


s 


.s 


UJ 




a 


03 


<U 


'£ 






tn 


X 

4) 


03 




a 


a 

OJ 


o 

OJ 


0) 


ro 




OJ 

> 


,s 


3 


> 


Ui 


Ui 




t/5 


>-i 


a 


3 


-a 


o3 


3 


"3 


U 


OJ 
1/2 


^ 


u 


"S 


r: 


3 


t/1 






c5 




OJ 
W) 

c3 


o3 
03 


_C 


o 


c 

03 


-3 


-a 

OJ 


OJ 


^^ 


JS 


aj 


J3 


H 




C 


E 

c4 


u 


'^ 


^ 


rt 




03 




a> 


3 


sz 


60 


d 






x: 


Si 



82 



A. H. DOERMANN 653 

(Fig. 3). The increase of infective centers in comparable cultures (Fig. 2) at 
these later stages is probably brought about by liberation of phage particles 
concurrent with this lysis. 

Experiments Using 5-Methyltryptophan as the Metabolic Inhibitor. — In trying 
to find a suitable metabolic inhibitor for instantaneously stopping phage 
growth, a large number of experiments was done using the antimetabolite 
5-methyltryptophan (5MT)- whose bacteriostatic action is blocked by trypto- 
phan (9). The technique used was similar to the cyanide lysis procedure except 
that tryptophan was omitted from the lysing medium and 5MT was used in 
place of cyanide. The results (?'ig. 4) are quite similar to the cyanide results in 
all respects except one. They are similar in failure to recover any phage particles 
during the early stages of the latent period, in the difiference between single and 
multiple infection, and in the shapes of the curves. They are different, however, 
in that both the single and the multiple infection curves are moved to the left 
along the time scale by 3 to 4 minutes. This indicates that more phage is liber- 
ated per cell if lysis is induced in the presence of 5MT than if it is brought about 
in the presence of CN~. This difference may be interpreted on the basis of two 
alternative hypotheses. 

First, it might be suspected that CN~ penetrates the cell and reaches its site 
of inhibition more quickly than 5MT. This would allow more phage repro- 
duction to go on between the time of exposure to the 5MT and the time at 
which the cell breaks open. In this event, a higher concentration of 5MT would 
enable penetration of an inhibitory amount in a shorter period of time, thus 
reducing the amount of phage found. To test this, the concentration of 5MT in 
the lysing medium was increased fivefold. No difference in the amount of phage 
liberated was found, suggesting that the rate of penetration of the poison is not 
limiting its effectiveness. 

A second hypothesis is that the reaction blocked by 5MT may be one of the 
earlier ones involved in the synthesis of phage constituents. At the time of 
addition of 5MT many individual phages may already have acquired these 
constituents and thus be able to go on to maturity before lysis disperses the 
enzyme equipment of the infected cell. Cyanide, on the other hand, may block 
one of the terminal reactions in phage production, with the result that at a 
given time fewer individuals will have passed this reaction than will have passed 
the 5MT-inhibitable step. Consequently, fewer particles will be liberated when 
using cyanide than when 5MT is used. 

DISCUSSION 

Earlier experiments (2) and tests made of the lysing efficiency of the T6 
stocks used here indicate that rapid lysis occurs when T6 stocks are added in 

^ Obtained through the courtesy of Dr. M. L. Tainter, Sterling-VVinthrop Research 
Institute, Rensselaer, New York. 



83 



654 INTRACELLULAR GROWTH OF BACTERIOPHAGES. I 

sufficient concentration to bacterial cultures. The very first experiments with 
bacteria infected with T5 (3) left no doubt that lysis from without by T6 will 
liberate T5 particles prematurely from infected bacteria. From the evidence 
contained in the present paper it cannot be definitely established whether the 
combined action of T6 and cyanide liberates all of the mature phage present in 
the cells. However, the fact that, during the terminal stages of intracellular 
development, the cyanide-lysis method yields as much phage as does spontane- 
ous lysis, suggests that the cyanide method liberates all of the mature phage. 
Furthermore, during the second half of the latent period, exactly the same 
amount of phage is liberated by cyanide alone as by cyanide plus T6. This sug- 
gests that cyanide acts promptly in arresting phage growth. Otherwise one 
would expect to find a consistently higher number of phage particles in the 
cyanide medium than in the medium in which cyanide and T6 are combined. 
The experiments presented here therefore warrant the working hypothesis that 
mature intracellular phage is effectively liberated by the treatment described, 
and that the method gives a true picture of the intracellular phage population. 
The validity of this working hypothesis will be conclusively demonstrated for 
the phage T3 in the second paper of this series (10). 

The bearing of the present experiments on our concept of phage reproduction 
might be discussed here. The finding that the original infecting particles are not 
recoverable from the cells during the first stages of the latent period appears at 
first sight surprising. Nevertheless some indirect evidence indicates that this is to 
be expected. The discovery that yields from mixedly infected bacteria may con- 
tain new combinations of the genetic material of the infecting types (11-13) 
suggests that some alteration of the infecting particles may occur. Furthermore, 
in mixed infections of bacteria with unrelated phages only one type is repro- 
duced. The other type, although adsorbed on the cells, it not only prevented from 
multiplying but the infecting particle of that type is lost (5, 14). On the basis of 
multiple infection experiments with ultraviolet-inactivated phage particles, 
Luria (15) has proposed that reproduction of phage occurs by reproduction of 
subunits which are at some later stage assembled into complete virus particles. 
The failure to find infective phage particles within the infected cell in the early 
stages of reproduction agrees with what would have been predicted from these 
experiments. 

The results of our experiments agree quite well with the scheme which 
Latarjet (16) suggested on the basis of x-ray inactivation studies of phage 
inside infected bacteria. Latarjet differentiated three segments of the latent 
period of phage growth. Using T2 he found that during the first segment of 6 to 
7 minutes' duration, singly infected bacteria show the same inactivation charac- 
teristics as do unadsorbed phage particles. In the second period, from time 7 to 
time 13 minutes, the phage in infected cells became more resistant to x-rays, 
even during the first 2 minutes of this segment in which the inactivation curves 



84 



A. H. DOERMANN 655 

Still retain a single hit character. During the last 4 minutes of this period the 
curves take on a multiple hit character. In the tinal segment, from time 13 
minutes to the end of the latent period, the curves retain the multiple hit charac- 
ter, but gradually regain the original x-ray sensitivity characteristic of free 
phage. These .\-ray experiments suggest again that a rather drastic alteration 
occurs to the infecting particle, and that particles with the original character- 
istics are not found in the cell until the second half of the latent period. This is 
precisely what is observed in the results presented here. Our experiments were 
done with T4, but comparison seems legitimate since the two viruses are quite 
closely related (17). 

Results of a similar nature to those discussed here were published by Foster 
(18). In studying the effect of proflavine on the growth of phage T2, Foster 
found that the time at which this poison was added influenced the amount of 
phage liberated by the bacteria. No phage is liberated from T2-infected cells 
(latent period 21 minutes) if proflavine is added during the first 12 minutes 
after infection even though lysis of the cells does occur at the normal time. 
When proflavine is added at later points in the latent period, lysis yields phage 
particles, the number depending on the time of proflavine addition. When the 
results of these single infection experiments are compared to the cyanide single 
infection experiments (Fig. 1) the results are seen to be quite similar. From 
other experiments Foster concluded that proflavine inhibits one of the final 
stages in the formation of fully infective phage. These facts, taken together, 
suggests that proflavine experiments were, in fact, measuring intracellular 
phage. 

SUMMARY 

A method is described for liberating and estimating intracellular bacterio- 
phage at any stage during the latent period by arresting phage growth and induc- 
ing premature lysis of the infected cells. This is brought about by placing the 
infected bacteria into the growth medium supplemented with 0.01 m cyanide 
and with a high titer T6 lysate. It was found in some of the later experiments 
that the T6 lysate is essential only during the first half of the latent period. 
Cyanide alone will induce lysis during the latter part of the latent period. 

Using this method on T4-infected bacteria it is found that during the first 
half of the latent period no phage particles, not even those originally infecting 
the bacteria, are recovered. This result is in agreement with the gradually 
emerging concept that a profound alteration of the infecting phage particle 
takes place before reproduction ensues. During the second half of the latent 
period mature phage is found to accumulate within the bacteria at a rate which 
is parallel to the approximately linear increase of intracellular DNA in this 
system. However, the phage production lags several minutes behind DNA 
production. 



85 



656 INTRACELLULAR GROWTH OF BACTERIOPHAGES. I 

When 5-methyltryptophan replaced cyanide as the metabolic inhibitor, 
similar results were obtained. The curves were, however, displaced several 
minutes to the left on the time axis. 

The results are compared with Latar jet's (16) data on x-radiation of infected 
bacteria and with Foster's data (18) concerning the effect of proflavine on in- 
fected bacteria. Essential agreement with both is apparent. 

BIBLIOGRAPHY 

1. Delbriick, M., J. Gen. Physiol, 1940, 23, 643. 

2. Doermann, A. H., /. Bad., 1948, 55, 257. 

3. Doermann, A. H., Ann. Rep. Biol. Lab., Long Island Biol. Assn., 1946, 22. 

4. Adams, M. H., J. Gen. Physiol, 1948, 31, 417. 

5. Delbriick, M., and Luria, S. E., Arch. Biochem., 1942, 1, 111. 

6. Cohen, S. S., Bad. Rev., 1949, 13, 1. 

7. Cohen, S. S., and Anderson, T. F., /. Exp. Med., 1946, 84, 525. 

8. Underwood, N., and Doermann, A. H., Rev. Scient. Instr., 1947, 18, 665. 

9. Anderson, T. F., Science, 1945, 101, 565. 

10. Anderson, T. F., and Doermann, A. H., /. Gen. Physiol, 1952, 35, 657. 

11. Delbruck, M., and Bailey, W. T., Jr., Cold Spring Harbor Symp. Qimnt. Biol, 

1946, 11, 33. 

12. Hershey, A. D., and Rotman, R., Proc. Nat. Acad. Sc, 1948, 34, 89. 

13. Hershey, A. D., and Rotman, R., Genetics, 1949, 34, 44. 

14. Delbruck, M., J. Bad., 1945, 50, 151. 

15. Luria, S. E., Proc. Nat. Acad-. Sc, 1947, 33, 253. 

16. Latarjet, R., /. Gen. Physiol, 1948, 31, 529. 

17. Adams, M. H., in Methods in Medical Research, (J. H. Comroe, editor), Chicago, 

The Yearbook Publishers, Inc., 1950, 2, 1. 

18. Foster, R. A. C, J. Bad., 1948, 56, 795. 



86 



INDEPENDENT FUNCTIONS OF VIRAL PROTEIN AND NUCLEIC 
ACID IN GROWTH OF BACTERIOPHAGE* 

By a. D. HERSHEY and MARTHA CHASE 

{From the Department of Genetics, Carnegie Institution of Washington, Cold Spring 

Harbor, Long Island) 

(Received for publication, April 9, 1952) 

The work of Doermann (1948), Doermann and Dissosway (1949), and 
Anderson and Doermann (1952) has shown that bacteriophages T2, T3, and 
T4 multiply in the bacterial cell in a non-infective form. The same is true of 
the phage carried by certain lysogenic bacteria (Lwoff and Gutmann, 1950). 
Little else is known about the vegetative phase of these viruses. The experi- 
ments reported in this paper show that one of the first steps in the growth of 
T2 is the release from its protein coat of the nucleic acid of the virus particle, 
after which the bulk of the sulfur-containing protein has no further function. 

Materials and Methods. — Phage T2 means in this paper the variety called T2H 
(Hershey, 1946); T2// means one of the host range mutants of T2; UV-phage means 
phage irradiated with ultraviolet light from a germicidal lamp (General Electric 
Co.) to a fractional survival of 10~^. 

Sensitive bacteria means a strain (H) of Escherichia coli sensitive to T2 and its 
h mutant; resistant bacteria B/2 means a strain resistant to T2 but sensitive to its 
h mutant; resistant bacteria B/2/; means a strain resistant to both. These bacteria 
do not adsorb the phages to which they are resistant. 

"Salt-poor" broth contains per liter 10 gm. bacto-peptone, 1 gm. glucose, and 1 
gm. NaCl. "Broth" contains, in addition, 3 gm. bacto-beef extract and 4 gm. NaCl. 

Glyceroldactate medium contains per liter 70 mM sodium lactate, 4 gm. glycerol, 
5 gm. NaCl, 2 gm. KCl, 1 gm. NH4CI, 1 mM MgCl2, 0.1 mM CaCl2, 0.01 gm. gelatin, 
10 mg. P (as orthophosphate), and 10 mg. S (as MgS04), at pH 7.0. 

Adsorption medium contains per liter 4 gm. NaCl, 5 gm. K2SO4, 1.5 gm. KH2PO4, 
3.0 gm. Na2HP04, 1 mM MgS04, 0.1 mM CaCU, and 0.01 gm. gelatin, at pH 7.0. 

Veronal buffer contains per liter 1 gm. sodium diethylbarbiturate, 3 mM MgS04, 
and 1 gm. gelatin, at pH 8.0. 

The HCN referred to in this paper consists of molar sodium cyanide solution 
neutralized when needed with phosphoric acid. 



* This investigation was supported in part by a research grant from the National 
Microbiological Institute of the National Institutes of Health, Public Health Service. 
Radioactive isotopes were supplied by the Oak Ridge National Laboratory on alloca- 
tion from the Isotopes Division, United States Atomic Energy Commission. 

Reprinted by permission of the authors and The Rockefeller 

Institute from The Journal of General Physiology, 36 (1), 

39-56, September 20, 1952. 

87 



40 VIRAL PROTEIN AND NUCLEIC ACID IN BACTERIOPHAGE GROWTH 

Adsorption of isotope to bacteria was usually measured by mixing the sample 
in adsorption medium with bacteria from 18 hour broth cultures previously heated 
to 70°C. for 10 minutes and washed with adsorption medium. The mixtures were 
warmed for 5 minutes at 37°C., diluted with water, and centrifuged. Assays were 
made of both sediment and supernatant fractions. 

Precipitation of isotope with antiserum was measured by mixing the sample in 
0.5 per cent saline with about 10" per ml. of non-radioactive phage and slightly 
more than the least quantity of antiphage serum (final dilution 1:160) that would 
cause visible precipitation. The mixture was centrifuged after 2 hours at 37°C. 

Tests with DNase (desoxyribonuclease) were performed by warming samples 
diluted in veronal buffer for 15 minutes at 37°C. with 0.1 mg. per ml. of crystalline 
enzyme (Worthington Biochemical Laboratory). 

Acid-soluble isotope was measured after the chilled sample had been precipitated 
with 5 per cent trichloroacetic acid in the presence of 1 mg./ml. of serum albumin, and 
centrifuged. 

In all fractionations involving centrifugation, the sediments were not washed, and 
contained about 5 per cent of the supernatant. Both fractions were assayed. 

Radioactivity was measured by means of an end-window Geiger counter, using 
dried samples sufficiently small to avoid losses by self-absorption. For absolute meas- 
urements, reference solutions of P^^ obtained from the National Bureau of Standards, 
as well as a permanent simulated standard, were used. For absolute measurements 
of S^^ we relied on the assays (±20 per cent) furnished by the supplier of the isotope 
(Oak Ridge National Laboratory). 

Glycerol-lactate medium was chosen to permit growth of bacteria without un- 
desirable pH changes at low concentrations of phosphorus and sulfur, and proved 
useful also for certain experiments described in this paper. 18-hour cultures of sensitive 
bacteria grown in this medium contain about 2 X 10^ cells per ml., which grow ex- 
ponentially without lag or change in light-scattering per cell when subcultured in 
the same medium from either large or small seedings. The generation time is 1.5 hours 
at 37°C. The cells are smaller than those grown in broth. T2 shows a latent period 
of 22 to 25 minutes in this medium. The phage yield obtained by lysis with cyanide 
and UV-phage (described in context) is one per bacterium at 15 minutes and 16 
per bacterium at 25 minutes. The final burst size in diluted cultures is 30 to 40 per 
bacterium, reached at 50 minutes. At 2 X 10* cells per ml., the culture lyses slowly, 
and yields 140 phage per bacterium. The growth of both bacteria and phage in this 
medium is as reproducible as that in broth. 

For the preparation of radioactive phage, P^^ of specific activity 0.5 mc./mg. or 
S^^ of specific activity 8.0 mc./mg. was incorporated into glycerol-lactate medium, 
in which bacteria were allowed to grow at least 4 hours before seeding with phage. 
After infection with phage, the culture was aerated overnight, and the radioactive 
phage was isolated by three cycles of alternate slow (2000 G) and fast (12,000 G) 
centrifugation in adsorption medium. The suspensions were stored at a concentration 
not exceeding 4 )uc./ml. 

Preparations of this kind contain 1.0 to 3.0 X 10-^^ ^g s and 2.5 to 3.5 X 10-" 
/xg. P per viable phage particle. Occasional preparations containing excessive amounts 
of sulfur can be improved by absorption with heat-killed bacteria that do not adsorb 



88 



A. D. HERSHEY AND MARTHA CHASE 



41 



the phage. The radiochemical purity of the preparations is somewhat uncertain, ow- 
ing to the possible presence of inactive phage particles and empty phage membranes. 
The presence in our preparations of sulfur (about 20 per cent) that is precipitated by 
antiphage serum (Table I) and either adsorbed by bacteria resistant to phage, or 
not adsorbed by bacteria sensitive to phage (Table VII), indicates contamination 
by membrane material. Contaminants of bacterial origin are probably negligible for 
present purposes as indicated by the data given in Table I. For proof that our prin- 
cipal findings reflect genuine properties of viable phage particles, we rely on some 
experiments with inactivated phage cited at the conclusion of this paper. 

The Chemical Morphology of Resting Phage Particles.— Anderson (1949) 
found that bacteriophage T2 could be inactivated by suspending the particles 
in high concentrations of sodium chloride, and rapidly diluting the suspension 
with water. The inactivated phage was visible in electron micrographs as tad- 
pole-shaped "ghosts." Since no inactivation occurred if the dilution was slow 

TABLE I 

Composition of Ghosts and Solution of Plasmolyzed Phage 



Per cent of isotope] 



Acid-soluble 

Acid-soluble after treatment with DNase 

Adsorbed to sensitive bacteria 

Precipitated by antiphage 



Whole phage labeled with 



pj2 



S»s 



1 

85 
90 



1 
90 
99 



Plasmolyzed 
phage labeled with 



P" 



1 

80 

2 
5 



S»5 



1 
90 
97 



he attributed the inactivation to osmotic shock, and inferred that the particles 
possessed an osmotic membrane. Herriott (1951) found that osmotic shock 
released into solution the DNA (desoxypentose nucleic acid) of the phage 
particle, and that the ghosts could adsorb to bacteria and lyse them. He pointed 
out that this was a beginning toward the identification of viral functions with 
viral substances. 

We have plasmolyzed isotopically labeled T2 by suspending the phage 
(10" per ml.) in 3 m sodium chloride for 5 minutes at room temperature, and 
rapidly pouring into the suspension 40 volumes of distilled water. The plas- 
molyzed phage, containing not more than 2 per cent survivors, was then an- 
alyzed for phosphorus and sulfur in the several ways shown in Table I. The 
results confirm and extend previous findings as follows: — 

1. Plasmolysis separates phage T2 into ghosts containing nearly all the 
sulfur and a solution containing nearly all the DNA of the intact particles. 

2. The ghosts contain the principal antigens of the phage particle detect- 
able by our antiserum. The DNA is released as the free acid, or possibly linked 
to sulfur-free, apparently non-antigenic substances. 



89 



42 



VIRAL PROTEIN AND NUCLEIC ACID IN BACTERIOPHAGE GROWTH 



3. The ghosts are speciJ&cally adsorbed to phage-susceptible bacteria; the 
DNA is not. 

4. The ghosts represent protein coats that surround the DNA of the intact 
particles, react with antiserum, protect the DNA from DNase (desoxyribo- 
nuclease), and carry the organ of attachment to bacteria. 

5. The effects noted are due to osmotic shock, because phage suspended in 
salt and diluted slowly is not inactivated, and its DNA is not exposed to 
DNase. 

TABLE II 

Sensitization of Phage DNA to DNase by Adsorption to Bacteria 



Phage adsorbed to 




Phage labeled 
with 


Non-sedimentable isotope, 
per cent 




After DNase 


No DNase 




§36 
p32 

s« 

p32 

S36 
p32 

p32 
p32 
P32 
p32 


2 
8 

15 
76 

12 
66 

5 
13 
81 
88 


1 


(( « 


7 


Bacteria heated before infection 


11 


11 K H >< 


13 


Bacteria heated after infection 


14 




23 


Heated unadsorbed phage: acid- 
soluble P« 


■ 70° 

80° 

90° 

100° 





Phage adsorbed to bacteria for 5 minutes at 37°C. in adsorption medium, followed by 
washing. 

Bacteria heated for 10 minutes at 80°C. in adsorption medium (before infection) or in 
veronal buffer (after infection). 

Unadsorbed phage heated in veronal buffer, treated with DNase, and precipitated with 
trichloroacetic acid. 

All samples fractionated by centrifuging 10 minutes at 1300 G. 

Sensitization of Phage DNA to DNase by Adsorption to Bacteria. — The struc- 
ture of the resting phage particle described above suggests at once the possibil- 
ity that multiplication of virus is preceded by the alteration or removal of the 
protective coats of the particles. This change might be expected to show itself 
as a sensitization of the phage DNA to DNase. The experiments described 
in Table II show that this happens. The results may be summarized as fol- 
lows: — 

1. Phage DNA becomes largely sensitive to DNase after adsorption to 
heat-killed bacteria. 

2. The same is true of the DNA of phage adsorbed to live bacteria, and then 



90 



A. D. HERSHEY AND MARTHA CHASE 43 

heated to 80°C. for 10 minutes, at which temperature unadsorbed phage is 
not sensitized to DNase. 

3. The DNA of phage adsorbed to unheated bacteria is resistant to DNase, 
presumably because it is protected by cell structures impervious to the en- 
zyme. 

Graham and collaborators (personal communication) were the first to dis- 
cover the sensitization of phage DNA to DNase by adsorption to heat-killed 
bacteria. 

The DNA in infected cells is also made accessible to DNase by alternate 
freezing and thawing (followed by formaldehyde fixation to inactivate cellu- 
lar enzymes), and to some extent by formaldehyde fixation alone, as illus- 
trated by the following experiment. 

Bacteria were grown in broth to 5 X 10^ cells per ml., centrifuged, resuspended in 
adsorption medium, and infected with about two P^Mabeled phage per bacterium. 
After 5 minutes for adsorption, the suspension was diluted with water containing per 
liter 1.0 mM MgS04, 0.1 mM CaCh, and 10 mg. gelatin, and recentrifuged. The cells 
were resuspended in the fluid last mentioned at a concentration of 5 X 10^ per ml. 
This suspension was frozen at -15°C. and thawed with a minimum of warming, 
three times in succession. Immediately after the third thawing, the cells were fixed by 
the addition of 0.5 per cent (v/v) of formalin (35 per cent HCHO). After 30 minutes 
at room temperature, the suspension was dialyzed free from formaldehyde and cen- 
trifuged at 2200 G for 15 minutes. Samples of P^^.j^beled phage, frozen-thawed, fixed, 
and dialyzed, and of infected cells fixed only and dialyzed, were carried along as 
controls. 

The analysis of these materials, given in Table III, shows that the effect 
of freezing and thawing is to make the intracellular DNA labile to DNase, 
without, however, causing much of it to leach out of the cells. Freezing and 
thawing and formaldehyde fixation have a negligible efifect on unadsorbed 
phage, and formaldehyde fixation alone has only a mild effect on infected cells. 

Both sensitization of the intracellular P^- to DNase, and its failure to leach 
out of the cells, are constant features of experiments of this type, independently 
of visible lysis. In the experiment just described, the frozen suspension cleared 
during the period of dialysis. Phase-contrast microscopy showed that the cells 
consisted largely of empty membranes, many apparently broken. In another 
experiment, samples of infected bacteria from a culture in salt-poor broth 
were repeatedly frozen and thawed at various times during the latent period 
of phage growth, fixed with formaldehyde, and then washed in the centrifuge. 
Clearing and microscopic lysis occurred only in suspensions frozen during the 
second half of the latent peritad, and occurred during the first or second thaw- 
ing. In this case the lysed cells consisted wholly of intact cell membranes, 
appearing empty except for a few small, rather characteristic refractile bodies 
apparently attached to the cell walls. The behavior of intracellular P^^ toward 
DNase, in either the lysed or unlysed cells, was not significantly different from 



91 



44 



VIRAL PROTEIN AND NUCLEIC ACID IN BACTERIOPHAGE GROWTH 



that shown in Table III, and the content of P^^ was only slightly less after 
lysis. The phage liberated during freezing and thawing was also titrated in 
this experiment. The lysis occurred without appreciable liberation of phage 
in suspensions frozen up to and including the 16th minute, and the 20 min- 
ute sample yielded only five per bacterium. Another sample of the culture 
formalinized at 30 minutes, and centrifuged without freezing, contained 66 
per cent of the P^^ in non-sedimentable form. The yield of extracellular phage 
at 30 minutes was 108 per bacterium, and the sedimented material consisted 
largely of formless debris but contained also many apparently intact cell 
membranes. 

TABLE III 

Sensitization of Intracellular Phage to DNase by Freezing, Thawing, and Fixation 
with Formaldehyde 



Unadsorbed 
phage frozen, 
thawed, fixed 



Infected cells 

frozen, thawed, 

fixed 



Infected cells 
fixed only 



Low speed sediment fraction 



Total P'2 


— 


71 



59 


86 


Acid-soluble 


5 


Acid-soluble after DNase 


28 





Low speed supernatant fraction 




Total P32 


1 


29 
0.8 
21 


Acid-soluble 


Acid-soluble after DNase. . 




11 



14 


0.4 


5.5 



The figures express per cent of total P'^ in the original phage, or its adsorbed fraction. 

We draw the following conclusions from the experiments in which cells 
infected with P^"-labeled phage are subjected to freezing and thawing. 

1. Phage DNA becomes sensitive to DNAse after adsorption to bacteria in 
buffer under conditions in which no known growth process occurs (Benzer, 
1952; Dulbecco, 1952). 

2. The cell membrane can be made permeable to DNase under conditions 
that do not permit the escape of either the intracellular P^- or the bulk of the 
cell contents. 

3. Even if the cells lyse as a result of freezing and thawing, permitting escape 
of other cell constituents, most of the P^^ derived from phage remains inside 
the cell membranes, as do the mature phage progeny. 

4. The intracellular P'^ derived from phage is largely freed during spon- 
taneous lysis accompanied by phage liberation. 



92 



A. D. HERSHEY AND MARTHA CHASE 



45 



We interpret these facts to mean that intracellular DNA derived from phage 
is not merely DNA in solution, but is part of an organized structure at all times 
during the latent period. 

Liberation of DNA from Phage Particles by Adsorption to Bacterial Frag- 
ments.— Tht sensitization of phage DNA to specific depolymerase by adsorp- 
tion to bacteria might mean that adsorption is followed by the ejection of 
the phage DNA from its protective coat. The following experiment shows that 
this is in fact what happens when phage attaches to fragmented bacterial 
cells. 

TABLE IV 
Release of DNA from Phage Adsorbed to Bacterial Debris 



Phage labeled with 



S" 



pj2 



Sediment fraction 



Surviving phage 

Total isotope 

Acid-soluble isotope 

Acid-soluble after DNase. 




22 

55 

2 

29 



Supernatant fraction 



Surviving phage 

Total isotope 

Acid-soluble isotope 

Acid-soluble after DNase. 




45 
0. 
39 



S''- and P52-labeled T2 were mixed with identical samples of bacterial debris in adsorption 
medium and warmed for 30 minutes at 37°C. The mixtures were then centrifuged for 15 
minutes at 2200 G, and the sediment and supernatant fractions were analyzed separately. 
The results are expressed as per cent of input phage or isotope. 

Bacterial debris was prepared by infecting cells in adsorption medium with 
four particles of T2 per bacterium, and transferring the cells to salt-poor 
broth at 37°C. The culture was aerated for 60 minutes, m/50 HCN was added, 
and incubation continued for 30 minutes longer. At this time the yield of ex- 
tracellular phage was 400 particles per bacterium, which remained unadsorbed 
because of the low concentration of electrolytes. The debris from the lysed 
cells was washed by centrifugation at 1700 G, and resuspended in adsorption 
medium at a concentration equivalent to 3 X 10^ lysed cells per ml. It consisted 
largely of collapsed and fragmented cell membranes. The adsorption of radio- 
active phage to this material is described in Table IV. The following facts 
should be noted. 



93 



46 VIRAL PROTEIN AND NUCLEIC ACID IN BACTERIOPHAGE GROWTH 

1. The unadsorbed fraction contained only 5 per cent of the original phage 
particles in infective form, and only 13 per cent of the total sulfur. (Much 
of this sulfur must be the material that is not adsorbable to whole bacteria.) 

2. About 80 per cent of the phage was inactivated. Most of the sulfur of 
this phage, as well as most of the surviving phage, was found in the sediment 
fraction. 

3. The supernatant fraction contained 40 per cent of the total phage DNA 
(in a form labile to DNase) in addition to the DNA of the unadsorbed surviving 
phage. The labile DNA amounted to about half of the DNA of the inactivated 
phage particles, whose sulfur sedimented with the bacterial debris. 

4. Most of the sedimentable DNA could be accounted for either as surviving 
phage, or as DNA labile to DNase, the latter amounting to about half the DNA 
of the inactivated particles. 

Experiments of this kind are unsatisfactory in one respect: one cannot tell 
whether the liberated DNA represents all the DNA of some of the inactivated 
particles, or only part of it. 

Similar results were obtained when bacteria (strain B) were lysed by large 
amounts of UV-killed phage T2 or T4 and then tested with P^Mabeled T2 
and T4. The chief point of interest in this experiment is that bacterial debris 
saturated with UV-killed T2 adsorbs T4 better than T2, and debris saturated 
with T4 adsorbs T2 better than T4. As in the preceding experiment, some of 
the adsorbed phage was not inactivated and some of the DNA of the inacti- 
vated phage was not released from the debris. 

These experiments show that some of the cell receptors for T2 are different 
from some of the cell receptors for T4, and that phage attaching to these spe- 
cific receptors is inactivated by the same mechanism as phage attaching to 
unselected receptors. This mechanism is evidently an active one, and not 
merely the blocking of sites of attachment to bacteria. 

Removal of Phage Coats from Infected Bacteria. — Anderson (1951) has ob- 
tained electron micrographs indicating that phage T2 attaches to bacteria 
by its tail. If this precarious attachment is preserved during the progress of 
the infection, and if the conclusions reached above are correct, it ought to 
be a simple matter to break the empty phage membranes off the infected 
bacteria, leaving the phage DNA inside the cells. 

The following experiments show that this is readily accomplished by strong 
shearing forces applied to suspensions of infected cells, and further that in- 
fected cells from which 80 per cent of the sulfur of the parent virus has been 
removed remain capable of yielding phage progeny. 

Broth-grown bacteria were infected with S"^- or P^^-labeled phage in ad- 
sorption medium, the unadsorbed material was removed by centrifugation, 
and the cells were resuspended in water containing per liter 1 mM MgS04, 
0.1 mM CaCl2, and 0.1 gm. gelatin. This suspension was spun in a Waring 



94 



A. D. HERSHEY AND MARTHA CHASE 



47 



blendor (semimicro size) at 10,000 r.p.m. The suspension was cooled briefly 
in ice water at the end of each 60 second running period. Samples were removed 
at intervals, titrated (through antiphage serum) to measure the number of 
bacteria capable of yielding phage, and centrifuged to measure the proportion 
of isotope released from the cells. 

The results of one experiment with each isotope are shown in Fig. 1. The 
data for S^^ and survival of infected bacteria come from the same experiment, 
in which the ratio of added phage to bacteria was 0.28, and the concentrations 



© 



Infected bactema 



© 




35 
® ExtpacellalQP 5 



Extracellulap P 



32 



Min. 



8 



2 3 4 

l^unninq time in blendop 

Fig. 1. Removal of S'^ and P^^ from bacteria infected with radioactive phage, and 
survival of the infected bacteria, during agitation in a Waring blendor. 

of bacteria were 2.5 X 10* per ml. infected, and 9.7 X 10^ per ml. total, by 
direct titration. The experiment with P^Mabeled phage was very similar. 
In connection with these results, it should be recalled that Anderson (1949) 
found that adsorption of phage to bacteria could be prevented by rapid stir- 
ring of the suspension. 

At higher ratios of infection, considerable amounts of phage sulfur elute 
from the cells spontaneously under the conditions of these experiments, though 
the elution of P^'" and the survival of infected cells are not affected by multi- 
plicity of infection (Table V). This shows that there is a cooperative action 
among phage particles in producing alterations of the bacterial membrane 
which weaken the attachment of the phage. The cellular changes detected in 



95 



48 



VIRAL PROTEIN AND NUCLEIC ACID IN BACTERIOPHAGE GROWTH 



this way may be related to those responsible for the release of bacterial com- 
ponents from infected bacteria (Prater, 1951; Price, 1952). 

A variant of the preceding experiments was designed to test bacteria at a 
later stage in the growth of phage. For this purpose infected cells were aerated 
in broth for 5 or 15 minutes, fixed by the addition of 0.5 per cent {v/v) com- 
mercial formalin, centrifuged, resuspended in 0.1 per cent formalin in water, 
and subsequently handled as described above. The results were very similar 
to those already presented, except that the release of P^^ from the cells was 
slightly less, and titrations of infected cells could not be made. 

The S^^-labeled material detached from infected cells in the manner de- 
scribed possesses the following properties. It is sedimented at 12,000 G, though 
less completely than intact phage particles. It is completely precipitated by 

TABLE V 

Eifect of Mulliplicily of Infection on Elution of Phage Membranes from Infected Bacteria 



Running time 


Multiplicity of 
infection 


P«-Iabeled phage 


S"-labeIed phage 


in blender 


Isotope eluted 


Infected bacteria 
surviving 


Isotope eluted 


Infected bacteria 
surviving 


min. 



2.5 



2.5 


0.6 
0.6 
6.0 
6.0 


per cent 

10 
21 
13 

24 


per cent 

120 
82 
89 
86 


per cent 
16 
81 
46 
82 


per cent 
101 

78 
90 
85 



The infected bacteria were suspended at 10' cells per ml. in water containing per liter 
1 mM MgSO^, 0.1 mM CaCla, and 0.1 gm. gelatin. Sampjes were withdrawn for assay of 
extracellular isotope and infected bacteria before and after agitating the suspension. In 
either case the cells spent about 15 minutes at room temperature in the eluting fluid. 

antiphage serum in the presence of whole phage carrier. 40 to 50 per cent of 
it readsorbs to sensitive bacteria, almost independently of bacterial concen- 
tration between 2 X 10^ and 10^ cells per ml., in 5 minutes at 37°C. The ad- 
sorption is not very specific: 10 to 25 per cent adsorbs to phage-resistant bac- 
teria under the same conditions. The adsorption requires salt, and for this 
reason the efficient removal of S^^ from infected bacteria can be accomplished 
only in a fluid poor in electrolytes. 

The results of these experiments may be summarized as follows: — ■ 

1. 75 to 80 per cent of the phage sulfur can be stripped from infected cells 
by violent agitation of the suspension. At high multiplicity of infection, nearly 
50 per cent elutes spontaneously. The properties of the S^^-labeled material 
show that it consists of more or less intact phage membranes, most of which 
have lost the ability to attach specifically to bacteria. 

2. The release of sulfur is accompanied by the release of only 21 to 35 per 



96 



A. D. HERSHEY AND MARTHA CHASE 49 

cent of the phage phosphorus, half of which is given up without any mechan- 
ical agitation. 

3. The treatment does ngt cause any appreciable inactivation of intracellu- 
lar phage. 

4. These facts show that the bulk of the phage sulfur remains at the cell 
surface during infection, and takes no part in the multiplication of intracellu- 
lar phage. The bulk of the phage DNA, on the other hand, enters the cell soon 
after adsorption of phage to bacteria. 

Transfer of Sulfur and Phosphorus from Parental Phage to Progeny. — We 
have concluded above that the bulk of the sulfur-containing protein of the 
resting phage particle takes no part in the multiplication of phage, and in 
fact does not enter the cell. It follows that little or no sulfur should be trans- 
ferred from parental phage to progeny. The experiments described below show 
that this expectation is correct, and that the maximal transfer is of the order 
1 per cent 

Bacteria were grown in glycerol-lactate medium overnight and subcultured 
in the same medium for 2 hours at 37°C. with aeration, the size of seeding 
being adjusted nephelometrically to yield 2 X 10^ cells per ml. in the sub- 
culture. These bacteria were sedimented, resuspended in adsorption medium 
at a concentration of 10^ cells per ml., and infected with S^^-labeled phage 
T2. After 5 minutes at 37°C., the suspension was diluted with 2 volumes of 
water and resedimented to remove unadsorbed phage (5 to 10 per cent by 
titer) and S^^ (about 15 per cent). The cells were next suspended in glycerol- 
lactate medium at a concentration of 2 X 10^ per ml. and aerated at 37°C. 
Growth of phage was terminated at the desired time by adding in rapid suc- 
cession 0.02 mM HCN and 2 X 10" UV-killed phage per ml. of culture. The 
cyanide stops the maturation of intracellular phage (Doermann, 1948), and 
the UV-killed phage minimizes losses of phage progeny by adsorption to bac- 
terial debris, and promotes the lysis of bacteria (Maal0e and Watson, 1951). 
As mentioned in another connection, and also noted in these experiments, 
the lysing phage must be closely related to the phage undergoing multiplica- 
tion {e.g., T2H, its h mutant, or T2L, but not T4 or T6, in this instance) in 
order to prevent inactivation of progeny by adsorption to bacterial debris. 

To obtain what we shall call the maximal yield of phage, the lysing phage 
was added 25 minutes after placing the infected cells in the culture medium, 
and the cyanide was added at the end of the 2nd hour. Under these condi- 
tions, lysis of infected cells occurs rather slowly. 

Aeration was interrupted when the cyanide was added, and the cultures 
were left overnight at 37°C. The lysates were then fractionated by centrifuga- 
tion into an initial low speed sediment (2500 G for 20 minutes), a high speed 
supernatant (12,000 G for 30 minutes), a second low speed sediment obtained 
by recentrifuging in adsorption medium the resuspended high speed sediment, 
and the clarified high speed sediment. 



97 



50 



VIRAL PROTEIN AND NUCLEIC ACID IN BACTERIOPHAGE GROWTH 



The distribution of S*^ and phage among fractions obtained from three cul- 
tures of this kind is shown in Table VI. The results are typical (except for the 
excessively good recoveries of phage and S^^) of lysates in broth as well as 
lysates in glycerol-lactate medium. 

The striking result of this experiment is that the distribution of S^^ among 
the fractions is the same for early lysates that do not contain phage progeny, 
and later ones that do. This suggests that little or no S^* is contained in the 
mature phage progeny. Further fractionation by adsorption to bacteria con- 
firms this suggestion. 

Adsorption mixtures prepared for this purpose contained about 5 X 10* 
heat-killed bacteria (70°C. for 10 minutes) from 18 hour broth cultures, and 

TABLE VI 

Per Cent Distributions of Phage and 5'^ among Centrifugally Separated Fractions of Lysates 
after Infection with S^^-Labeled T2 



Fraction 


Lysis at 
/ = 

S3S 


Lysis at 

I = 10 

S35 


Maximal yield 




S35 


Phage 


1st low speed sediment 

2nd " " " 


79 
2.4 
8.6 

10 


81 
2.1 
6.9 

10 


82 
2.8 

7.1 

7.5 


19 

14 


Hieh speed " 


61 


" " supernatant 


7.0 


Recovery 


100 


100 


96 


100 



Infection with S'*-labeled T2, 0.8 particles per bacterium. Lysing phage UV-killed h 
mutant of T2. Phage yields per infected bacterium: <0.1 after lysis at / = 0; 0.12 at / = 
10; maximal yield 29. Recovery of S'^ means per cent of adsorbed input recovered in the 
four fractions; recovery of phage means per cent of total phage 3'ield (by plaque count before 
fractionation) recovered by titration of fractions. 

about 10'' phage (UV-killed lysing phage plus test phage), per ml. of adsorp- 
tion medium. After warming to 37°C. for 5 minutes, the mixtures were diluted 
with 2 volumes of water, and centrifuged. Assays were made from supernatants 
and from unwashed resuspended sediments. 

The results of tests of adsorption of S^'^ and phage to bacteria (H) adsorbing 
both T2 progeny and //-mutant lysing phage, to bacteria (B/2) adsorbing lysing 
phage only, and to bacteria (B/2//) adsorbing neither, are shown in Table VII, 
together wdth parallel tests of authentic S^^-labeled phage. 

The adsorption tests show that the S^^ present in the seed phage is adsorbed 
with the specificity of the phage, but that S^^ present in lysates of bacteria 
infected with this phage shows a more complicated behavior. It is strongly 
adsorbed to bacteria adsorbing both progeny and lysing phage. It is weakly 
adsorbed to bacteria adsorbing neither. It is moderately well adsorbed to bac- 



98 



A. D. HERSHEY AND MARTHA CHASE 



51 



teria adsorbing lysing phage but not phage progeny. The latter test shows that 
the S^^ is not contained in the phage progeny, and explains the fact that the 
S^^ in early lysates not containing progeny behaves in the same way. 

The speciticity of the adsorption of S^^-labeled material contaminating the 
phage progeny is evidently due to the lysing phage, which is also adsorbed 
much more strongly to strain H than to B/2, as shown both by the visible re- 
duction in Tyndall scattering (due to the lysing phage) in the supernatants 
of the test mixtures, and by independent measurements. This conclusion is 
further confirmed by the following facts. 

TABLE VII 

Adsorption Tests with Uniformly S^^-Labeled Phage and with Products of Their Growth in 

N on-Radioactive Medium 





Per cent adsorbed 


Adsorbing bacteria 


Uniformly labeled 
S35 phage 


Products of 
lysis at 
t = 10 


Phage progeny 
(Maximal yield) 




+ UV-A 


No UV-A 




S35 


S3S 


S" 

79 
46 
29 


S35 


Phage 


Sensitive (H) .... .... 


84 

15 
13 


86 
11 
12 


78 
49 

28 


96 


Resistant (B/2) 


10 


Resistant (B/2A) 


8 







The uniformly labeled phage and the products of their growth are respectively the seed 
phage and the high speed sediment fractions from the experiment shown in Table VI. 

The uniformly labeled phage is tested at a low ratio of phage to bacteria: -{-\]V-h means 
with added UV-killed h mutant in equal concentration to that present in the other test 
materials. 

The adsorption of phage is measured by plaque counts of supernatants, and also sedi- 
ments in the case of the resistant bacteria, in the usual way. 

1. If bacteria are infected with S" phage, and then lysed near the midpoint 
of the latent period with cyanide alone (in salt-poor broth, to prevent read- 
sorption of S^^ to bacterial debris), the high speed sediment fraction contains 
S^^ that is adsorbed weakly and non-specifically to bacteria. 

2. If the lysing phage and the S^^-labeled infecting phage are the same 
(T2), or if the culture in salt-poor broth is allowed to lyse spontaneously (so 
that the yield of progeny is large), the S^^ in the high speed sediment fraction 
is adsorbed with the specificity of the phage progeny (except for a weak non- 
specific adsorption). This is illustrated in Table VII by the adsorption to H 
and B/2/?. 

It should be noted that a phage progeny grown from S^Mabeled phage and 
containing a larger or smaller amount of contaminating radioactivity could 
not be distinguished by any known method from authentic S^^-labeled phage, 



99 



52 VIRAL PROTEIN AND NUCLEIC ACID ]N BACTERIOPHAGE GROWTH 

except that a small amount of the contaminant could be removed by adsorp- 
tion to bacteria resistant to the phage. In addition to the properties already- 
mentioned, the contaminating S^* is completely precipitated with the phage 
by antiserum, and cannot be appreciably separated from the phage by further 
fractional sedimentation, at either high or low concentrations of electrolyte. 
On the other hand, the chemical contamination from this source would be 
very small in favorable circumstances, because the progeny of a single phage 
particle are numerous and the contaminant is evidently derived from the 
parents. 

The properties of the S^Mabeled contaminant show that it consists of the 
remains of the coats of the parental phage particles, presumably identical 
with the material that can be removed from unlysed cells in the Waring blen- 
dor. The fact that it undergoes little chemical change is not surprising since it 
probably never enters the infected cell. 

The properties described explain a mistaken preliminary report (Hershey 
et al., 1951) of the transfer of S^^ from parental to progeny phage. 

It should be added that experiments identical to those shown in Tables VI 
and VII, but starting from phage labeled with P^-, show that phosphorus is 
transferred from parental to progeny phage to the extent of 30 per cent at 
yields of about 30 phage per infected bacterium, and that the P^- in prema- 
turely lysed cultures is almost entirely non-sedimentable, becoming, in fact, 
acid-soluble on aging. 

Similar measures of the transfer of P^'^ have been published by Putnam and 
Kozlofif (1950) and others. Watson and Maaljzie (1952) summarize this work, 
and report equal transfer (nearly 50 per cent) of phosphorus and adenine. 

A Progeny of S^^-Labeled Phage Nearly Free from the Parental Label. — -The 
following experiment shows clearly that the obligatory transfer of parental 
sulfur to offspring phage is less than 1 per cent, and probably considerably 
less. In this experiment, the phage yield from infected bacteria from which 
the S^Mabeled phage coats had been stripped in the Waring blendor was 
assayed directly for S^^. 

Sensitive bacteria grown in broth were infected with five particles of S^^-Iabeled 
phage per bacterium, the high ratio of infection being necessary for purposes of as- 
say. The infected bacteria were freed from unadsorbed phage and suspended in water 
containing per liter 1 mM MgS04, 0.1 mM CaCl2, and 0.1 gm. gelatin. A sample of 
this suspension was agitated for 2.5 minutes in the Waring blendor, and centrifuged 
to remove the extracellular S^\ A second sample not run in the blendor was centri- 
fuged at the same time. The cells from both samples were resuspended in warm 
salt-poor broth at a concentration of 10^ bacteria per ml., and aerated for 80 min- 
utes. The cultures were then lysed by the addition of 0.02 mM HCN, 2 X 10" UV- 
killed T2, and 6 mg. NaCl per ml. of culture. The addition of salt at this point causes 
S^^ that would otherwise be eluted (Hershey et al., 1951) to remain attached to the 



100 



A. D. HERSHEY AND MARTHA CHASE 



53 



bacterial debris. The lysates were fractionated and assayed as described previously, 
with the results shown in Table VIII. 

The data show that stripping reduces more or less proportionately the S^*- 
content of all fractions. In particular, the S^^-content of the fraction contain- 
ing most of the phage progeny is reduced from nearly 10 per cent to less than 
1 per cent of the initially adsorbed isotope. This experiment shows that the 
bulk of the S^* appearing in all lysate fractions is derived from the remains of 
the coats of the parental phage particles. 

Properties of Phage Inactivated by Formaldehyde. — Phage T2 warmed for 
1 hour at 37°C. in adsorption medium containing 0.1 per cent {v/v) com- 
mercial formalin (35 per cent HCHO), and then dialyzed free from formalde- 

TABLE VIII 
Lysates of Bacteria Infected with S^^-Labeled T2 and Stripped in the Waring Blendor 



Per cent of adsorbed S'' or of phage yield: 



Elated in blendor fluid. 
1st low-speed sediment. 
2nd " " " . 

High-speed " 

" " supernatant . 



Recovery 93 



Cells stripped 



S« 



Phage 



Cells not stripped 



S" 



86 
3.8 
(0.2) 
(0.7) 
(2.0) 



9.3 
11 
58 

1.1 



79 



39 

31 
2.7 
9.4 

(1.7) 



84 



Phage 



13 
11 
89 
1.6 



115 



All the input bacteria were recovered in assays of infected cells made during the latent 
period of both cultures. The phage yields were 270 (stripped cells) and 200 per bacterium, 
assayed before fractionation. Figures in parentheses were obtained from counting rates 
close to background. 

hyde, shows a reduction in plaque titer by a factor 1000 or more. Inactivated 
phage of this kind possesses the following properties. 

1. It is adsorbed to sensitive bacteria (as measured by either S^^ or P^^ 
labels), to the extent of about 70 per cent. 

2. The adsorbed phage kills bacteria with an efficiency of about 35 per 
cent compared with the original phage stock. 

3. The DNA of the inactive particles is resistant to DNase, but is made 
sensitive by osmotic shock. 

4. The DNA of the inactive particles is not sensitized to DNase by adsorp- 
tion to heat-killed bacteria, nor is it released into solution by adsorption to 
bacterial debris. 

5. 70 per cent of the adsorbed phage DNA can be detached from infected 
cells spun in the Waring blendor. The detached DNA is almost entirely re- 
sistant to DNase. 



101 



54 VIRAL PROTEIN AND NUCLEIC ACID IN BACTERIOPHAGE GROWTH 

These properties show that T2 inactivated by formaldehyde is largely in- 
capable of injecting its DNA into the cells to which it attaches. Its behavior in 
the experiments outlined gives strong support to our interpretation of the cor- 
responding experiments with active phage. 

DISCUSSION 

We have shown that when a particle of bacteriophage T2 attaches to a 
bacterial cell, most of the phage DNA enters the cell, and a residue contain- 
ing at least 80 per cent of the sulfur-containing protein of the phage remains 
at the cell surface. This residue consists of the material forming the protective 
membrane of the resting phage particle, and it plays no further role in infec- 
tion after the attachment of phage to bacterium. 

These facts leave in question the possible function of the 20 per cent of sul- 
fur-containing protein that may or may not enter the cell. We find that little 
or none of it is incorporated into the progeny of the infecting particle, and that 
at least part of it consists of additional material resembling the residue that 
can be shown to remain extracellular. Phosphorus and adenine (Watson and 
Maal0e, 1952) derived from the DNA of the infecting particle, on the other 
hand, are transferred to the phage progeny to a considerable and equal ex- 
tent. We infer that sulfur-containing protein has no function in phage multi- 
plication, and that DNA has some function. 

It must be recalled that the following questions remain unanswered. (1) 
Does any sulfur-free phage material other than DNA enter the cell? (2) If 
so, is it transferred to the phage progeny? (3) Is the transfer of phosphorus (or 
hypothetical other substance) to progeny direct — that is, does it remain at 
all times in a form specifically identifiable as phage substance — or indirect? 

Our experiments show clearly that a physical separation of the phage T2 
into genetic and non-genetic parts is possible. A corresponding functional 
separation is seen in the partial independence of phenotype and genotype in 
the same phage (Novick and Szilard, 1951; Hershey el al., 1951). The chemi- 
cal identification of the genetic part must wait, however, until some of the 
questions asked above have been answered. 

Two facts of significance for the immunologic method of attack on problems 
of viral growth should be emphasized here. First, the principal antigen of the 
infecting particles of phage T2 persists unchanged in infected cells. Second, it 
remains attached to the bacterial debris resulting from lysis of the cells. These 
possibilities seem to have been overlooked in a study by Rountree (1951) of 
viral antigens during the growth of phage T5. 

SUMMARY 

1. Osmotic shock disrupts particles of phage T2 into material containing 
nearly all the phage sulfur in a form precipitable by antiphage serum, and 
capable of specific adsorption to bacteria. It releases into solution nearly all 



102 



A. D. HERSHEY AND MARTHA CHASE 55 

the phage DNA in a form not precipitable by antiserum and not adsorbable 
to bacteria. The sulfur-containing protein of the phage particle evidently 
makes up a membrane that protects the phage DNA from DNase, comprises 
the sole or principal antigenic material, and is responsible for attachment of 
the virus to bacteria. 

2. Adsorption of T2 to heat-killed bacteria, and heating or alternate freezing 
and thawing of infected cells, sensitize the DNA of the adsorbed phage to 
DNase. These treatments have little or no sensitizing effect on unadsorbed 
phage. Neither heating nor freezing and thawing releases the phage DNA 
from infected cells, although other cell constituents can be extracted by these 
methods. These facts suggest that the phage DNA forms part of an organized 
intracellular structure throughout the period of phage growth. 

3. Adsorption of phage T2 to bacterial debris causes part of the phage 
DNA to appear in solution, leaving the phage sulfur attached to the debris. 
Another part of the phage DNA, corresponding roughly to the remaining half 
of the DNA of the inactivated phage, remains attached to the debris but can 
be separated from it by DNase. Phage T4 behaves similarly, although the 
two phages can be shown to attach to different combining sites. The inactiva- 
tion of phage by bacterial debris is evidently accompanied by the rupture of 
the viral membrane. 

4. Suspensions of infected cells agitated in a Waring blendor release 75 per 
cent of the phage sulfur and only 15 per cent of the phage phosphorus to the 
solution as a result of the applied shearing force. The cells remain capable of 
yielding phage progeny. 

5. The facts stated show that most of the phage sulfur remains at the cell 
surface and most of the phage DNA enters the cell on infection. Whether 
sulfur-free material other than DNA enters the cell has not been determined. 
The properties of the sulfur-containing residue identify it as essentially un- 
changed membranes of the phage particles. All types of evidence show that 
the passage of phage DNA into the cell occurs in non-nutrient medium under 
conditions in which other known steps in viral growth do not occur. 

6. The phage progeny yielded by bacteria infected with phage labeled 
with radioactive sulfur contain less than 1 per cent of the parental radioactiv- 
ity. The progeny of phage particles labeled with radioactive phosphorus con- 
tain 30 per cent or more of the parental phosphorus. 

7. Phage inactivated by dilute formaldehyde is capable of adsorbing to 
bacteria, but does not release its DNA to the cell. This shows that the inter- 
action between phage and bacterium resulting in release of the phage DNA 
from its protective membrane depends on labile components of the phage 
particle. By contrast, the components of the bacterium essential to this inter- 
action are remarkably stable. The nature of the interaction is otherwise un- 
known. 

8. The sulfur-containing protein of resting phage particles is confined to a 



103 



56 VIRAL PROTEIN AND NUCLEIC ACID IN BACTERIOPHAGE GROWTH 

protective coat that is responsible for the adsorption to bacteria, and functions 
as an instrument for the injection of the phage DNA into the cell. This pro- 
tein probably has no function in the growth of intracellular phage. The DNA 
has some function. Further chemical inferences should not be drawn from the 
experiments presented. 

REFERENCES 

Anderson, T. F., 1949, The reactions of bacterial viruses with their host cells, Bot. 
Rev., 15, 464. 

Anderson, T. F., 1951, Tr. New York Acad. Sc, 13, 130. 

Anderson, T. F., and Doermann, A. H., 1952, /. Gen. Physiol., 35, 657. 

Benzer, S., 1952, /. Bad., 63, 59. 

Doermann, A. H., 1948, Carnegie Institution of Washington Yearbook, No. 47, 176. 

Doermann, A. H., and Dissosway, C, 1949, Carnegie Institution of Washington Year- 
book, No. 48, 170. 

Dulbecco, R., 1952, /. Bad., 63, 209. 

Harriott, R. M., 1951, /. Bad., 61, 752. 

Hershey, A. D., 1946, Gettdics, 31, 620. 

Hershey, A. D., Roesel, C, Chase, M., and Forman, S., 1951, Carnegie Institution of 
Washington Yearbook, No. 50, 195. 

Lwoff, A., and Gutmann, A., 1950, Ann. Inst. Pasteur, 78, 711. 

Maal0e, O., and Watson, J. D., 1951, Proc. Nat. Acad. Sc, 37, 507. 

Novick, A., and Szilard, L., 1951, Science, 113, 34. 

Prater, C. D., 1951, Thesis, University of Pennsylvania. 

Price, W. H., 1952, J. Gen. Physiol, 35, 409. 

Putnam, F. W., and Kozloff, L.,1950, /. Biol. Chem., 182, 243. 

Rountree, P. M., 1951, Brit. J. Exp. Path., 32, 341. 

Watson, J. D., and Maal0e, O., 1952, Acta path, et microbial, scand., in press. 



104 



NUCLEIC ACID TRANSFER FROM PARENTAL 
TO PROGENY BACTERIOPHAGE 

by 

J. D. WATSON*. ** AND O. MAAL0E 

Institute for Cytophysiology and State Serum Institute, Copenhagen {Denmark) 



INTRODUCTION 

In recent years several investigators have used isotopic markers to determine the 
transfer of P or N atoms from parental to progeny virus particles. Putnam and KozloffI 
using the bacteriophage T6 labelled with ^^V, found that 20-40% of the label appeared 
in material identified as progeny phage by differential centrifugation. Leslie et al.^ 
found similar, or lower, values with ^^P-labelled T2 bacteriophage. Kozloff^ extended 
his original observations by studying P and N transfer from normal as well as radiation 
damaged phage particles. All these experiments show incomplete transfer varying 
greatly from one experiment to the other. 

To determine the true transfer values, two technical problems must be solved: 
(i) adsorption of the labelled phage must be complete within about 2 minutes because 
later adsorbing particles are broken down before entering the cells (Leslie et al.^), 
(2) means must be found to prevent progeny particles from adsorbing onto bacterial 
debris or unlysed cells. Failure to control either of these processes will result in under- 
estimation of the transfer values. 

In our first experiments (Maal0e and Watson*), although the second factor was 
well controlled, the first was not, and our ^^F transfer values of about 30% were, there- 
fore, hke those of Putnam and KozloffI. Leslie et al.^, and Kozloff^, too low. In 
this paper, we present experiments in which both factors are controlled, and which 
indicate that T2 and T4 phages transfer about 50%, T3 phages about 40% of their 
phosphorus to the progeny. Identical values are found when, instead of ^^p, i*C-labelled 
adenine is used to label the parental phage. The ^2? experiments show that the trans- 
ferred material goes predominantly to the early formed phages, and that the transfer 
values of about 50% are maximum values. Our finding that the early completed phages 
receive most of the parental material confirms similar observations by Doermann 
(personal communication) and Weed and Cohen^ (unpubhshed) . 

In agreement with Kozloff^, we have also demonstrated that considerable amounts 
of 32? may be transferred from labelled particles which do not participate in repro- 
duction, either because of radiation damage or because they are excluded by another 
phage. These "abnormal" cases show that infecting particles may be broken down 

* Merck Fellow in the Natural Sciences of the National Research Council at the Institute for 
Cytophysiology, Copenhagen. 

** Present address; Cavendish Laboratory, Cambridge, England. 

References p. 442. 

Reprinted by permission of the authors and Elsevier Publishing Co. 
from BiocmMiCA et Biophysica Acta, 10, 432-442 (1953). 

105 



VOL. 10 (1953) NUCLEIC ACID TRANSFER 433 

extensively beforeHheir constituents are used for synthesis of new phage particles. We 
do not, however, think that this evidence excludes the possibility that under normal 
conditions the infecting and reproducing particle remains essentially unbroken. We shall 
return to this important point in the discussion. 

MATERIALS AND METHODS 

The phages T2, T3 andT4, and their common host E. coli, strain B/i, have been used: the latter 
because contamination of cultures with phage Tr occurs in our laboratory. Most of the techniques 
used in this study have been described in detail by Adams^. All cultures were grown at 37° C: cen- 
trifugations were done in a Servall Angle centrifuge at 10° C. 

Media. The nutrient broth is an aqueous extract of minced meat enriched with i % peptone 
and containing 0.02% Tween 80 and 0.5% NaCl: pH is adjusted to 7.4. For experiments with T3, 
the concentration of NaCl was reduced to 0.05 % to obtain rapid adsorption. In this medium, 
at 37° C the latent period of T4r is 22-23 minutes and the burst size about 150. ^^p-iabelled phage 
was prepared using the synthetic g-medium described earlier (Maaloe and Watson*). 

Anti-sera. Rabbit anti-phage sera were prepared, using highly purified and concentrated phage 
suspensions as antigens. All sera had k-values (Adams^) of 500-1000 when tested against phage 
suspended in broth (of. Jerne'). These sera showed no agglutination of E. coli, strain B/i, in dilution 
T to 10. Samples of anti-phage sera were absorbed with large amounts of live B/i, and used in parallel 
with unabsorbed serum for phage precipitation : no evidence for the presence of antibacterial anti- 
bodies was found. Serum against E. coli. strain B/i, was obtained after series of subcutaneous and 
intravenous injections of heat-killed and subsequently of live cultures. This serum had no anti-phage 
activity. 

X-ray technique. The X-ray source was a Holbeck-Beaudouin tube operating at 33 kv and 
36 mA. A cooled molybdenum target produced radiation with an average wave-length of 0.9 A. 
On the surface of the samples the intensity of radiation was 66,000 r.p.m. The irradiation was done 
by Dr R. Latarjet on samples sent to Paris by airmail. No decrease in titer was observed in control 
samples as a result of the shipment. 

Isotope technique. Carrier-free orthophosphoric acid was obtained from the Isotope Division 
of the Oak Ridge National Laboratory, United States Atomic Energy Commission. Adenine labelled 
with "C in position 8 was also used: this preparation was synthesized by Clark and Kalckar*, 
and had a specific activity of 0.8 mC per mMole. All samples containing "C were evaporated to 
dryness and self-absorption due to solids in the suspension medium was made uniform by diluting 
into nutrient broth before counting. Variation between counts on duplicate ^^C samples was less 
than 10%. The counting equipment was the same as previously described. 

Preparation of ^^P-labelled T4r. Washed B/i bacteria from a 24-hour broth culture were in- 
oculated into 10 ml of g-medium containing 20 uC ^^p. After two and a half hours of aeration, the 
bacterial density was about lo^ cells per ml, and the culture was then infected with about 10 T4r 
particles per cell Aeration was continued, and 2-3 minutes before the onset of lysis 0.5 ml of un- 
diluted antibacterial serum was added. If antibacterial serum is not added, the titer of a crude 
T4r lysate will drop appreciably during the first 24 hours and a fraction of the remaining phages 
will adsorb slowly. Both effects are presumably due to phage particles absorbing on bacterial debris 
(Maaloe and Stent^). The antibacterial serum proved completely effective in blocking adsorption 
of T3 and T4 on B/i, but was not fully effective for T2r+. T2r+ stocks may, therefore contain inactive 
as well as slowly adsorbing particles even when antibacterial serum is used. It is possible to restore 
infectivity and full adsorbability to these particles by diluting into distilled water for several hours 
at 37° C: presumably because adsorbed particles dissociate from the debris at low salt concentrations 
(Puck, Garen, and ClineI" and Hershey, personal communication). This treatment was first used 
by Bertani (personal communication) to raise the titer of T2r+ broth lysates. 

The crude radioactive lysates were centrifuged at 5000 g for 5 minutes to remove bacterial 
debris and at 12,000 g for one hour to sediment the phage. Three cycles of low and high speed 
centrifugation reduced the concentration of inorganic ^-P by a factor of about 10*. Further puri- 
fication was achieved by adding heat-killed resistant bacteria (B/3, 4, 7 heated to 58° C for i hour) 
at a concentration of 5- lo^ per ml. After 30 minutes at 37° C, about 5% of the radioactivity had 
adsorbed to the resistant cells, which were removed by centrifugation. A similar number of sensitive 
B/i cells adsorbed 95-98% of the activity. An additional test of the purity of the virus stock was 
obtained by precipitation with antiphage serum, as previously described; with anti-T4 serum, 94% 
of the radioactivity was precipitated, while in a control sample in which T3 phage was precipitated 
with anti-T3 serum, the precipitate contained less than 2% of the activity. 

Assuming that about 95 % of the ^^P in the final preparation was present as phage phosphorus, 
the initial specific activity was io~^ counts per minute per particle. From this we can calculate 

References p. 442. 

106 



434 



J. D. WATSON, O. MAAL0E VOL. 10 (1953) 



that each virus particle contained an average of 0.25 ^zp atoms. Since the inactivation efficiency 
of nuclear decay is only 1/12 (Hershey, Kamen, Kennedy and Gest^^), the fraction of labelled 
particles which will lose infectivity during an experiment is negligible. 

With minor variation indicated by adsorption requirements, etc.. the procedure just described 
was also used to prepare ^^p-iabelled stocks of the phages T2r+, T^, and T4r+. Phage labelled with 
i*C adenine was grown on a purine requiring mutant of E. coli, strain B, which was obtained from 
Dr A. H. Doermann of the Oak Ridge National Laboratory. For this purpose the g-mediura was 
supplemented with 5 y "C adenine per ml. The specific activity of phages grown on the purine 
requiring strain was ten times that of phages grown similarly on B/i. 



EXPERIMENTAL 

The basic experimental procedure is unchanged throughout this study. It will, 
therefore, suffice to describe one typical and simple experiment in detail ; more complex 
experiments can then be introduced briefly and the results summarized in tables. 

Distribution of ^^P following infection of unlabelled Bfi with labelled T4r : 

Exponentially growing bacteria from an unlabelled broth culture were collected 
by centrifugation and resuspensed in unlabelled broth at 37° C at a concentration of 
2-io9 cells per ml. ^ap-labelled T4r phage was added at a ratio of 4.5 particles per bac- 
terium and one and one-half minutes allowed for adsorption. The culture was then 
chilled and centrifuged at 5000 g for 5 minutes. The supernatant was carefully siphoned 
off and samples for phage assay and radioactivity measurement were taken. The pellet 
was resuspended in broth at 37° C to give a suspension of infected bacteria with about 
108 cells/ml; this figure was determined by a separate assay. Aeration was then started, 
and 25 minutes after infection one volume of undiluted antibacterial serum was added 
to 19 volumes of culture to prevent adsorption of progeny phage particles to bacterial 
debris. The cooling and centrifugation retarded phage growth by about 8 minutes. 
The assays and radioactivity measurements showed that in this experiment over 99.5% 
of the phages and 96% of the input radioactivity were adsorbed on the bacteria. 

About 30 minutes after lysis, the culture was centrifuged at 5000 g for 5 minutes 
to remove bacterial debris and then at 12,000 g for one hour to sediment the progeny 
phage. The material collected during these centrifugations will be referred to as the 
"low speed pellet" and the "high speed pellet", respectively. The latter was resuspended 
in broth and again centrifuged at low speed to remove remaining bacterial debris. The 

TABLE I 

distribution of 32p AFTER INFECTION OF B/l WITH LABELLED T4r 

Growing bacteria were concentrated to 2-108 cells per ml and infected with labelled T4r at 
a concentration of 9- 10* particles per ml. Following an adsorption period of 1 1/2 minutes the bacteria 
were centrifuged for 4 minutes at 5000 g to remove unadsorbed phage and then resuspended m 
nutrient broth at a concentration of li-io^ cells per ml. Approximately 99-5% of the phage and 
96% of ^2P adsorbed to the bacteria. 





Material 


Phage titer Imi 


^^P distribution 
(% of radioactivity 
adsorbed on bacteria) 




Crude lysate 
Low Speed Pellets 
High Speed Pellet 
High Speed Supernatant 


I.5-IO10 
.027- lO^" 

1.45 •10"' 

.II-IO'" 


100% 

8.4% 
42.1% 
49-5% 


References p. 442. 









107 



VOL. 10 (1953) NUCLEIC ACID TRANSFER 435 

two low speed pellets were resuspended in broth and all fractions assayed for ^^p and 
phage. Table I shows that 42.1% of the ^^p which initially adsorbed to the bacteria 
and about 95% of the progeny phage was recovered in the high speed pellet. 

The radioactivity of the high speed pellet was characterized as belonging to phage 
particles by precipitation with an excess of anti-T4 serum and by adsorption to sensitive 
bacteria. Native as well as B/i adsorbed serum was used and in both cases 93% of 
the ^2p was precipitated with the phage. Adsorption tests showed that over 90% of 
the radioactivity adsorbed on B/i cells while less than 5% adsorbed on the resistant 
strain B/3, 4, 7 on which T4 does not itself adsorb. These tests show that 90-95% of 
the radioactivity in the high speed pellet is somehow associated with the phage particles. 

Before asserting that this ^^P is truly incorporated into the progeny phages, the 
following possibihties must be considered: (i) degraded parental nucleic acids might 
stay attached to the surface of the progeny particles, or (2) non-infective parental 
particles might be adsorbed to the bacteria initially, released during lysis, and later 
sediment together with the progeny particles. 

Both these possibihties are ruled out by the fact that the progeny particles from 
an experiment like the one just described transfer their ^^P in exactly the same way 
as did their uniformly labelled parents (Maaloe and Watson*). We, therefore, conclude 
that about 95% of the ^^P in the high speed pellet is incorporated into the progeny 
particles. Using this estimate in correcting for the phage present in the low speed pellets 
and in the high speed supernatant, the fraction of the parental ^^p which has become 
incorporated into the progeny particles is 45% in this experiment. 

Table II shows a series of similar experiments involving different stocks of labelled 
T4r; it is notable that the differences in transfer are ve'"y small, usually less than 5%. 
These values are not changed by varying the number of infective particles from i to 
10 (the highest number tested). The amount of ^^P remaining attached to bacterial 
debris is uniformly about 5 to 10%. If the different phage stocks had contained greatly 
varying fractions of adsorbing but non-infecting phages, this would have caused the 
low speed pellet values to fluctuate greatly. 

TABLE II 

EXPERIMENTAL VARIATION IN ^'P TRANSFER VALUES FOR T4r 





Burst size 


% 


of parental radioactivity 


in 


Experiment 


Low speed 
pellets 


Progeny phage 


High speed 
supernatant 


I 


122 


6 


47 


47 


2 


141 


7 


46 


47 


3 


150 


6 


45 


49 


4 


126 


3 


45 


52 


5 


160 


9 


43 


48 


6 


155 


10 


42 


48 



^^P transfer from secondarily adsorbed phage 

The experiments given in Table III show that particles which adsorb on a bacterium 
more than two minutes after a primary infection with T4r transfer insignificant amounts 
of ^^P. This was demonstrated by infecting bacteria with an average of 5 non-labelled 
phages per cell and at various times later reinfecting with labelled phage. Column i 
References p. 442. 

108 



^36 J- D. WATSON, O. MAAL0E VOL. 10 (1953) 

of Table III further shows that with increasing intervals between primary and secondary 
infection, the amount of ^^p which stays attached to the bacteria decreases. Since the 
labelled phages adsorb at the normal rate, most of the ^^p in the supernate of the 
infected cells must initially have been adsorbed onto the cells. The observed increase 
in unadsorbed material is probably an expression of the breakdown of secondarily 
adsorbed particles described by Leslie, French, Graham, and Van Rooyen^. These 
authors observed that a primary infection stimulates within a few minutes the infected 
cell in such a way that, if a new phage particle adsorbs, it is broken down extensively 
on the surface of the bacterium, releasing about 50% of its phosphorus into the medium 
in the form of material soluble in 5 % trichloroacetic acid. 

This stimulation phenomenon is of further importance for transfer experiments, 
since we must assume that early released progeny particles which adsorb on unlysed 
cells will be broken down and release half their phosphorus into the medium. In most 
of the previous work adsorption of progeny particles was not prevented and the break- 
down effect presumably decreased the transfer values. Since in our experiments ad- 
sorption was blocked by antibacterial serum, the 50% transfer values can be taken as 
a good estimate of the transfer to all the progeny particles. 

TABLE III 

DISTRIBUTION OF ^^P FOLLOWING SECONDARY INFECTION OF B/l BY LABELLED T^T 

Growing bacteria concentrated to 1.3-10® cells/ml in broth were infected at i = o, with an 
average of 5 unlabelled T4r particles. At various intervals, labelled T4r was added at a ratio of 
I particle per bacterium. Several minutes after the addition of the labelled phage, unadsorbed ^^P 
was removed by low speed centrifugation and the infected bacteria resuspended in nutrient broth 
at a concentration of 10* cells/ml. 



Minutes between 

primary and secondary 

infection 



% of parental radioactivity in 



05 



Unadsorbed 

32p 


Low speed 
pellet 


Progeny phage 


High speed 
supernatant 


5 


6 


44 


45 


6 


4 


48 


42 


23 


15 


18 


44 


48 


23 


4 


25 


45 


20 


3 


32 



* This material must initially have been adsorbed but has been released again because of the 
breakdown effect described by Lesley et al.^. 

Isotope transfer from parental to progeny Tj 

At the salt concentration usually employed in growth media, T3 phage adsorbs 
rather slowly. To ensure the necessary rapid adsorption of the labelled T3 particles, 
the NaCl concentration in the adsorption tube had to be lowered to 0.05%. At this 
low salt concentration, over 99% of the infecting particles adsorbed within two minutes. 
Except for this modification, the experiments were like the T4 experiments and included 
characterization of the progeny by serum precipitation and adsorption on sensitive 
bacteria. Table IV shows that about 40% of the parental ^ap was transferred. Since 
nearly simultaneous adsorption of the infecting phage particles and isolation of the 
entire progeny were achieved, we may consider the transfer value of about 40% as a 
maximum value. As in the case of the phages T2 and T4, the transfer of phosphorus 
is incomplete. 
References p. 442. 

109 



VOL. 10 {1953) NUCLEIC ACID TRANSFER 437 

TABLE IV 

DISTRIBUTION OF 32p FOLLOWING INFECTION OF B/l BY LABELLED T3 





Burst size 


% 


of parental radioactivity 


' in 


Experiment 


Low speed 
pellet 


Progeny phage 


High speed 
supernatant 


I 
2 


195 
234 


9 
8 


46 
38 


45 
54 



Purine transfer from parental to progeny phage 

Over 95% of the phage phosphorus is located in DNA, and it is therefore desirable 
to know whether other nucleic acids constituents such as the purine bases are transferred 
incompletely like the phosphorus. Phages T2, T3 and T4 were grown in purine requiring 
bacteria in the presence of adenine labelled with ^*C in position 8 (see page 433 et seq.). 
Paper chromatography shows that in this way both the phage adenine and guanine is 
labelled with ^*C. Table V presents a series of experiments with the purine labelled 
phages. They all show incomplete transfer with values not significantly different from 
those obtained with ^^p ; X3 again seems to transfer a little less than do T2 and T4. 

TABLE V 

DISTRIBUTION OF ^*C FOLLOWING INFECTION OF B/l WITH PURINE LABELLED PHAGE 





Phage 


Burst size 


% of parent radioactivity in 


Experiment 


Low speed 


Progeny 


High speed 








pellet 


phage 


supernatant 


I 


T2r+ 


322 


12 


55 


33 


2 


T2r+ 


238 


16 


54 


30 


3 


T2r+ 


162 


18 


43 


39 


4 


T2r+ 


360 


14 


48 


38 


5 


T3 


180 


14 


38 


48 


6 


T3 


276 


II 


38 


51 


7 


T3 


210 


14 


32 


54 


8 


T4r 


115 


12 


44 


44 


9 


T4r 


122 


14 


40 


46 



Isotope transfer as a function of hurst size 

In experiments with T3 and T4r spontaneous lysis always occurs when the burst 
size is relatively low. It is known, however, that the r+-phages behave differently, 
and that lysis can be delayed by a secondary infection (Doermann^^^. At the time of 
normal lysis the infected cells have not exhausted their capacity for phage production. 
During the last ten minutes before normal lysis, large amounts of phosphorus-containing 
phage material is produced in the cells which is not developed into mature phage 
before lysis (Maal0e and Stent^). After normal lysis, this material cannot be recovered 
by centrifugation. The incompleteness of the phosphorus and purine transfers might, 
therefore, be due to our failure to detect parental material transmitted to the immature 
phage particles formed late in the latent period. 

This hypothesis was tested by infecting bacteria with labelled T4r+-phage and 
inhibiting lysis artificially by reinfection with unlabelled phage. The burst size in this 
References p. 442. 

110 



^38 J- D. WATSON, O. MAAL0E VOL. 10 {1953) 

experiment was 350, or 2 or 3 times greater than in the experiment with T4r. Despite 
this increase in phage yield, the transfer was again about 50%, showing that the late 
formed particles received very little, if any, of the parental phosphorus. 

The experiment reported in Table VI shows directly that the transferred ^2? goes 
predominantly to the early formed particles. In this experiment, bacteria infected with 
labelled T2r+-phage were lysed prematurely by addition of M/iooo KCN and a large 
number of ultraviolet inactivated phage, as previously described*. The excess of added 
phage served as carrier material during centrifugation, and insvued that the labelled 
progeny particles were effectively isolated even when the yield was low. Several identical 
experiments were carried out, all showing the same trend. Thus the material which an 
infecting particle transfers to the progeny usually goes to one or more of the early 
formed particles, while the later formed ones virtually never receive any of it. We 
therefore conclude that our transfer values of about 50% are true maximum values. 
Table VI also shows that bacteria lysed before the appearance of the first progeny 
particles contain insignificant amounts of radioactivity sedimentable at high speed. 
This is additional evidence that our results are not affected by spurious measurements 
of non-infective but still sedimentable parental particles released upon lysis. 

TABLE VI 

DISTRIBUTION OF PARENTAL 32p AMONG PROGENY PARTICLES FROM PREMATURELY LYSED BACTERIA 

Concentrated B/i were infected with an average of 5 ^^P-iabelled T2r+ particles per bacterium. 
Two minutes after infection the culture was chilled and centrifuged at 5000 g for 4 minutes to remove 
unadsorbed radioactivity. The pellet was resuspended in broth at 37° C at a bacterial concentration 
of 108 cells/ml. The progress of phage growth was retarded about 8 minutes by the coolmg and cen- 
trifugation. Ten minutes after infection ultraviolet (UV) inactivated T2r+ at the average multiplicity 
of 5 particles was added to inhibit lysis. Samples of the infected bacteria were then broken open 
at various times by the addition of Mjiooo KCN and approximately 2000 UV inactivated T2r+ 
particles per cell. Readsorption of the progeny particles was prevented by saturation of the bacterial 
surface with the UV treated phage (Maaloe and Watson*). 





Burst size 




% 


of parental ^^P in 




Time of premature 
lysis 


Low speed 




High speed 


High speed 






pellet 




pellet 


supernatant 


19 minutes 





13 




2 


85 


25 


34 


15 




15 


70 


28 


100 


14 




26 


60 


34 


210 


15 




38 


47 


48 


275 


16 




37 


47 



Isotope transfer in the absence of genetic transfer 

The "second generation experiment" (Maal0E and Watson*) referred to earher 
in this paper permits the conclusion that the transmitted phosphorus is distributed in 
the progeny particles in the same uniform way as in the parental particles. It leaves 
open the question whether transfer occurs via large blocks carrying biological specificity 
or via highly degraded material. In this study we have tried to answer a complementary 
question: Does isotope transfer occur under conditions where no genetic transfer is 
possible? i.e., can material from an infecting particle be broken down to genetically 
unspecific structures which are then incorporated into new phage particles? 
References p. 442. 

Ill 



VOL. 10 (1953) NUCLEIC ACID TRANSFER 



439 



TABLE VII 

TRANSFER OF ^2? FROM X-RAY INACTIVATED T4r 

Bacteria were mixedly infected with unlabelled T4r+ and X-ray irradiated 32p.iabelled T4r. 
The adsorption of ^zp to bacteria was similar in the mixtures containing irradiated phage and in 
control mixtures containing non-irradiated phage, being approximately 93 % in both cases. 



% of parental ^^P in 



Tube contents 



Low speed Progeny High speed 
pellet phage supernatant 



Unlabelled T4r+ 

+ labelled T4r (non-irradiated) 11 43 46 

Unlabelled T4r+ 

+ labelled T4r (e-^ survival) 31 24 45 

Unlabelled T4r+ 

+ labelled T4r (e-* survival) 45 22 33 

Unlabelled T4r+ 

+ labelled T4r (e-i" survival) 42 30 28 



To answer this question we first carried out experiments with labelled phage 
heavily irradiated with X-rays. Such particles retain the ability to adsorb to bacteria 
while the majority of them have lost not only their infectivity, but also the ability to 
kill the host cell and to transfer genetic specificity (WatsonI^). The experiments pre- 
sented in Table VII show, however, that heavy X-ray damage reduces the transfer 
values only from the normal 40 to 50% to about 25%. Thus, substantial amounts of 
phosphorus may be transferred from particles which do not participate in genetic ex- 
change. It should be noted that nearly 50% of the phosphorus of the irradiated particles 
remain attached to the bacterial debris. This value is significantly greater than the 
values of 5 to 15% found when active phage reproduces. It is possible that as many 
as 50 % of the irradiation damaged phages remains passively attached to the cell surfaces ; 
if so, the transfer value per transferring particle is again about 50%. 

Transfer of parental isotope without simultaneous genetic transfer can also be 
demonstrated in bacteria in which one of the infecting phages does not multiply because 
of the presence of an unrelated phage. This "mutual exclusion" (DelbruckI") is well 
illustrated by the unrelated phages T3 and T4; if a bacterium is infected simultaneously 
by both phages, T3 multipHcation is completely suppressed and only T4 progeny 
particles appear; even the infecting T3 particles are lost. In Table VIII are shown the 
results of an experiment in which bacteria were simultaneously infected with ^^P-Iabelled 
T3 particles and unlabelled T4 particles. After lysis, the progeny particles were isolated 
and tested for radioactivity, the specificity of which was shown by the extent of its 
precipitation with anti-T3 and anti-T4 serum. It can be seen that approximately 25 % 
of the 32p originally present in the T3 particles was transferred to the T4 progeny, 
another 25% was associated with bacterial debris, while the remaining 50% cannot be 
sedimented at high speed. The excluded phage thus does not sit passively on the bacterial 
surface but must penetrate to the interior of the cell where it is broken down into its 
simpler components. Our experiment, therefore, supports the conclusion of Weigle 
AND Delbruck^^ that mutual exclusion must involve some mechanism other than the 
establishment of a barrier to penetration. 
References p. 442. 

112 



440 



J. D. WATSON, O. MAAL0E VOL. 10 (1953) 



TABLE VIII 

TRANSFER OF '^P FROM EXCLUDED PHAGE 

Bacteria were mixedly infected with an average of 5 particles of T4r and i particle of labelled 
T3 per bacterium. Following lysis the progeny particles were isolated and tested for radioactivity. 





T3 


Burst 


size 
T4r 




% of parental (Tj) radioactivity in 




Experiment 

No. 


Low speed 
pellet 


High speed 
pellet 


% of high speed pellet radioactivity 
precipitated by 


High speed 
superruUant 




anti Tj serum anti T4 serum 


I 
2 

3 


2 

4 
5 




145 
162 
148 


20 
24 
19 


21 
27 
16 


4 92 


59 
49 
65 



DISCUSSION 

We shall now discuss the results of the seven different types of experiments de- 
scribed in the preceding section in relation to the problem of virus reproduction. First 
it must be stressed, however, that in all our experiments only the nucleic acid portion 
of the phage has been labelled ; entirely different results are obtained if a specific protein 
labeUing isotope, like ^^S, is used, (Hershey and Chase^^). 

1. Infection of unlabelled bacteria with ^sp-labelled T4r phage results in the transfer 
of 40 to 50% of the label to the progeny particles. The transmitted phosphorus is truly 
incorporated into the new phage. When the infected bacteria lyse, 5 to 10% of the 
parental ^^p remain associated with the bacterial debris; this may mean that 5 to 10% 
of the adsorbing particles stay passively attached to the bacterial surfaces. In our 
experiments the transfer value is an extremely reproducible figure and must be close 
to the maximum value since (a) nearly all the adsorbed particles participate in the 
reproduction process, and (b) all the progeny particles are recovered. The transfer value 
is constant for different preparations of labelled phage. 

2. For maximum ^ap transfer to occur, all the phage particles which adsorb on a 
bacterium must do so within about two minutes. Particles which adsorb more than 
two minutes after the primary infection of the cell do not transfer significant amounts 
of 32p. Half the ^^p of such late adsorbing particles remains attached to the bacterial 
debris; the rest appears within a few minutes in the medium as material soluble in 5% 
trichloroacetic acid. In contrast, over 95% of the ^^p of the early adsorbing phage is 
retained in the infected cells until lysis. Under conditions of almost simultaneous 
infection, the transfer value per infecting particle is constant for multipHcities of infection 
up to ten. 

3. 32p-iabelled T2 phage shows the same transfer as T4, and, according to Putnam 
AND KozLOFpi and Kozloff^, similar maximum values are obtained with the related 
phage T6. The unrelated phage T3 transfers about 40% of its phosphorus to the progeny. 
Thus an incomplete ^^P transfer of 40 to 50% may be a general characteristic of phage 
reproduction in E. coli. 

4. If the purines of T2, T3 and T4 phages are labelled with i*C, transfer values are 
obtained which do not differ from those obtained with ^^P. This suggests that the purine 
bases and the phosphorus are not transferred independently but as the constituents 
of nucleotides or larger units. 

References p. 442. 

113 



VOL. 10 (1953) NUCLEIC ACID TRANSFER 44I 

5. Most of the transmitted ^^p is incorporated into the early finished phage. An 
increase by a factor 2 to 3 in the phage yield per infected cell can be obtained in lysis 
inhibited cultures ; such an increase in yield does not significantly increase the transfer. 

6. The "second generation experiments" previously published (Maal0e and 
Watson*) show that uniformly labelled parental particles transmit phosphorus to all 
parts of the progeny particles which in turn become uniformly labelled. 

7. Phage heavily damaged by X-rays or excluded from growth by the simultaneous 
presence of an unrelated phage transfer part of their nucleic acid. Isotope transfer is, 
therefore, not necessarily connected with the transfer of genetic specificity. 

It is well known that infecting particles cannot be recovered by artificial lysis at 
any time during the first half of the latent period (Doermann"), and it is now firmly 
established that the nucleic acid components, phosphorus, adenine and guanine, are 
incompletely transferred from the infecting particle to the progeny. The transfer experi- 
ments, and especially the observation that chemical transfer may occur in the absence 
of genetic transfer, might, therefore, be viewed as evidence for an obligate and extensive 
breakdown of the infecting particle soon after adsorption. There is no doubt that such 
a breakdown can occur as evidenced by the unspecific transfer from radiation damaged 
or excluded phage (7). In both these cases, however, transfer takes place under abnormal 
conditions where the transferring particle does not participate in the reproduction 
processes. It is therefore unwarranted to conclude from these experiments that phos- 
phorus is always transferred via genetically unspecific structures. 

As they have turned out, the existing data on nucleic acid transfer, including the 
second generation experiments, do not decide whether transfer occurs via extensively 
degraded parental material or via large, genetically specific units. The data even fit 
with the assumption that nucleic acid structure of the infecting phage remains intact 
during replication and with a probability of about 50% becomes infective again and 
reappears among the progeny. 

ACKNOWLEDGEMENTS 

This investigation has been generously assisted by funds from the Danish Society 
for Infantile Paralysis and by the William Waterman Fund for the Combat of Dietary 
Diseases. The synthesis of the ^^C-adenine by Drs. M. Clark and H. M. Kalckar was 
made possible by a grant to Dr Kalckar from the Lederle Laboratories Division of 
the American Cyanamid Company. We are also grateful to Dr A. H. Doermann for 
supplying us with the purine requiring strain of E. coli, to Dr Dean Fraser for a most 
generous gift of anti-T3 rabbit serum, and to Dr R. Latarjet for putting the X-ray 
facilities of the Laboratoire Pasteur de ITnstitut du Radium, Paris, at our disposal. 



SUMMARY 

a. When ^^p-iabelled phage reproduce in unlabelled coli bacteria a maximum of 40-50% of 
their label is transmitted to the phage progeny. Only 5-10% of the label stay associated with bacterial 
debris after lysis: the remaining about 40% appear as non-sedimentable material in the lysate. 

b. The maximum transfer-values are very reproducible provided that all phages adsorbing on 
a given cell do so within about 2 minutes, and that the entire progeny is accounted for. 

c. Experiments with T2, T3 and T4 all show a maximum *^P transfer of 40-50%. The same 
phages labelled with i*C in the purines yield identical transfer-values. 

References p. 442. 

114 



442 J- D. WATSON, O. MAAL0E VOL. 10 {1953) 

d. The transmitted '^P is found predominantly in the early formed phages. The latest formed 
progeny particles receive no ^^p from the parental particles. 

e. Damaged or excluded phage particles which do not participate in reproduction or in genetic 
exchange nevertheless transmit considerable amounts of ^^P to the phage progeny. 

RfiSUMfi 

a. Lorsque des phages marques par ^P se reproduisent dans des colibacteries non marqu6es, 
le 40-50% au maximum de leur ^^P est transmis aux phages nouveaux. Le 10% du ^^P seulement 
teste dans les debris bacteriens apres la lyse: le reste, 40% environ, apparait comma matiere non 
sedimentable dans le lysat. 

b. Les valeurs maxima de transfert sont bien reproductibles pourvu que I'adsorption de tous 
les phages d'une meme cellule ait lieu en moins de 2 minutes environ et que Ton tienne compte 
de tous les phages nouveaux. 

c. Des experiences faites avec T2, T3 etT4 donnent toutes, pour le transfert de ^^P, un maximum 
de 40-50 % . Les memes phages, marques par ^*C dans les purines, donnent des valeurs de transfert 
identiques. 

d. Le ^^P transmis se trouve surtout dans les phages formes les premiers. Les phages de la 
nouvelle generation formes les derniers ne re9oivent pas de ^^P des particules meres. 

e. Des particules de phages endommagees ou "exclues" (c.a d., dont la croissahce est empechee 
par un autre phage) qui ne prennenb pas part a la reproduction ou a I'echange genetique, transmettent 
tout de meme des quantites considerables de ^^P a la nouvelle generation. 

ZUSAMMENFASSUNG 

a. Bei der Vermehrung von mit ^^P markierten Phagen in nicht markierten Colibakterien wird 
ein Maximum von 40-50% des markierten Phosphors auf die Phagennachkommenschaft iibertragen. 
Nur 5-10% des markierten Phosphors bleiben mit den Bakterienresten nach der Lysis verbunden. 
die iibrigen 40% erscheinen als nicht sedimentierbares Material im Lysat. 

b. Die Maximumiibertragungswerte sind sehr gut reproduzierbar, vorausgesehen, dass die 
Adsorption bei alien Phagen einer gegebenen Zelle innerhalb von zwei Minuten stattfindet und dass 
die gesamte Nachkommenschaft beriicksichtigt wird. 

c. Versuche mit T2, T3 und T4 zeigen alle ein Maximum der ^^p.tJbertragung von 40-50%. 
Die gleichen, mit ^*C in den Purinen markierten Phagen ergeben identische tlbertragungswerte. 

d. Der iibertragene ^^p findet sich iiberwiegend in den zuerst gebildeten Phagen wieder. Die 
zuletzt gebildeten Nachkommenteilchen erhalten keinen ^-P von den Elternteilchen. 

e. Beschadigte oder von der Vermehrung ausgeschlossene Phagenteilchen, die nicht an der 
Vermehrung oder dem genetischen Austausch teilnehmen, iibertragen trotzdem betrachtliche Mengen 
32p an die Phagennachkommenschaft. 

REFERENCES 

1 F. W. Putnam and L. Kozloff, /. Biol. Chem., 182 (1950) 243. 

2 S. M. Leslie, R. C. French, A. F. Graham and C. E. van Rooyen, Can. J. Med. Sci., 29 (1951) 128. 

3 L. M. Kozloff, /. Biol. Chem., 194 (1952) 95- 

* O. Maaloe and J. D. Watson, Proc. Natl. Acad. Sci.. 37 (1951) 507- 
^ L. Weed and S. S. Cohen, 195 i, unpublished. 

8 M. H. Adams, In Methods in Medical Research, Vol. ii, Year Book PubUshers. Chicago, 1950. 
' N. K. Jerne, Nature, 169 (1952) 117. 

8 M. M. Clark and H. M. Kalckar, /. Chem. Soc, (1950) 1029. 
^ O. Maaloe and G. S. Stent, Acta Path. Microbiol. Scand., (in press). 
10 J J Puck, A. Garen and J. Cline, /. Exptl Med., 93 (1951) 65. 

^1 A. D. Hershey, M. D. Kamen, J. W. Kennedy and H. Gest, /. Gen. Physiol., 23 (1951) 643. 
^2 A. H. Doermann, /. Bad., 55 (1948) 257. 
" J. D. Watson, /. Bad., 60 (1950) 697. 
^* M. Delbruck, /. Bad., 50 (1945) 137, 151. 
^5 J. J. Weigle and M. Delbruck, /. Bad., 62 (1951) 301. 
" A. D. Hershey and M. Chase, J. Gen. Phvsiol., in press. 
1^ A. H. Doermann, Carnegie Inst. Year Book, 47 (1948) 176. 

Received April 24th, 1952 



115 



THE SYNTHESIS OF BACTERIAL VIRUSES* 

II. THE ORIGIN OF THE PHOSPHORUS FOUND IN THE DESOXYRIBO- 
NUCLEIC ACIDS OF THE T2 AND T4 BACTERIOPHAGES 

By SEYMOUR S. COHEN 

(From the Children's Hospital of Philadelphia (Department of Pediatrics) and the 

Department of Phijsiological Chemistry, School of Medicine, 

University of Pennsylvania, Philadelphia) 

(Received for publication, August 28, 1947) 

It has been shown that Escherichia coli B infected by T2 bacteriophage in 
a synthetic medium, F, synthesizes protein-bound constituents containing 
phosphorus and nitrogen at a constant rate (1). Evidence was presented 
that suggests that the virus-infected cell synthesizes virus constituents 
solely. This was most clearly indicated in the phosphorus metabolism, 
since the only phosphorylated protein-bound constituent synthesized in the 
infected cell was that characteristic of virus; i.e., desoxyribonucleic acid 
(DNA). The ribonucleic acid (RNA) content of infected cells did not 
increase. 

Although the data presented suggested that the materials formed during 
infection were derived from the medium as in normal cells, it had not been 
proved that the nucleic acid phosphorus found in virus was not derived from 
cellular substance existing prior to infection, or had not been in normal host 
constituents during infection prior to appearance in virus. It seemed un- 
likely that the virus nucleic acid was derived from the host DNA, since sev- 
eral times more DNA was synthesized after infection than was originally 
present. However, it was possible that the RNA of the infected cell pos- 
sessed an active turnover yielding its phosphorus to DNA nucleotides, while 
the RNA was continually replaced with nucleotides whose phosphorus was 
derived from the medium. 

Two basic problems were therefore posed. (1) Was the DNA-P of virus 
derived from the phosphorus in the host prior to infection? (2) Was RNA 
a precursor of DNA? These hypotheses were tested by means of radio- 
active phosphorus, P^'. 

EXPERIMENTAL 

Conditions for Optimal Virus Yield — A single generation of virus multipli- 
cation was studied under conditions of multiple infection. The possibility 
was therefore eliminated that the generations of virus produced after the 
first generation in a medium of known composition would be affected by the 

*Thp work described in this paper was aided by the Office of Naval Research. 

Reprinted by permission of the author and the American Society 

of Biological Chemists, Inc. from The Journal of Biological 

Chemistry, 174 (1), 295-303 (1948). 

116 



296 BACTERIAL VIRUSES. II 

appreciable amounts of complex products, radioactive and otherwise, liber- 
ated by lysed cells. Some of the chemical phenomena and basic biological 
aspects of this system have been described in Paper I (1). 

The yield of virus from infected cells under conditions of multiple infec- 
tion in the aerated F medium at 37° was studied at concentrations of 5 X 
10^ 10^ 2 X 10^ and 5 X 10^ bacteria per cc. It was found that with both 
T2 and T4 maximal titers were obtained with 2 X lO'^ bacteria per cc. in 
about 4 to 6 hours. After this time, titers generally decreased markedly, 
and to a greater extent in T2 lysates than in T4 lysates. This was due 
probably to a combination of several factors, readsorption to cellular debris, 
spontaneous thermal inactivation, surface denaturation, etc. In view of 
this marked loss of titer, it was considered desirable to prepare virus lysates 
by a single generation in infected cells for a 6 hour period, to minimize the 
destruction of virus activity. 

Distribution of DMA and Protein-Bound P in To and Ti Lysates — Two 
types of experiments were done. (1) Bacteria were grown in media contain- 
ing radioactive P, washed several times, and infected in media containing 
non-radioactive P; (2) bacteria were grown in media free of P^^ and in- 
fected in the presence of P^-. 

The following typical control experiments were performed to indicate 
whether (1) the experimental conditions employed would permit the isola- 
tion of sufficient virus for analytical purposes, (2) virus isolated by these 
procedures would be very low in inorganic P and otherwise possess the 
proper chemical properties, (3) the newly synthesized DNA appeared in the 
virus fraction. Bacteria were grown to 2 X 10* per cc. in F medium. Two 
125 cc. ahquots were sedimented and the bacteria washed twice with 0.85 
per cent NaCl. Each aliquot was resuspended in 125 cc. portions of F me- 
dium to which were added an adsorption cofactor, 12 cc. of 5 X 10~^ m tryp- 
tophan ill F (2). To these bacterial suspensions (A and C) were added 
small volumes of purified concentrates of T4r+-F and T2r+-F to give cultures 
T4A and T2C containing 6 X 10* virus particles per cc. The virus concen- 
trates were also added to unwashed bacterial cultures (B and D) in F me- 
dium at the same concentrations to yield T4B and T2D; B contained 5 X 
10"^ M tryptophan. The four infected cultures were assayed periodically; 
the titers followed the course described previously. 

After 5 hours at 37°, the T2 and T4 lysates were stored for 13 hours at 4°. 
Very little inactivation occurred; the final titers were T4A 2.0 X 10^", 
T4B 4.0 X 1010, T2C 1.5 X 101", T2D 1.5 X 1010 active virus particles per 
cc. The lysates were analyzed for total protein-bound P and DNA (1). 
They were sedimented at 4000 r.p.m. for 30 minutes and the sediments were 
washed twice with cold 0.85 per cent NaCl. The supernatant fluids were 
sedimented at 10,000 r.p.m. for 2 hours, and the supernatant fluids from 



117 



S. S. COHEN 



297 



this high speed centrifugation, containing more than 5 per cent of the 
activity of the lysate, were analyzed for protein-bound P and DNA. Virus 
was isolated from the high speed sediments as described previously (3, 4). 
The distribution of these substances in the four lysates is given in Table I. 
It may be seen that the largest amounts of DNA and P appear in the 
virus-containing fraction; i.e., the high speed sediment. When this fraction 
was resuspended in 5 cc. of 0.85 per cent NaCl and centrifuged at 4000 
R.p.M. for 30 minutes, approximately 25 per cent of the DNA of the fraction 
had become insoluble. After dialysis of this supernatant fluid against 
running water overnight, the once sedimented virus concentrate had 
DNA-P to total P ratios which were very close to 1.0. These values for 
isolated dialyzed virus after one differential centrifugation cycle were 

Table I 
Distribution of Protein-Bound P and DNA in Lysates of Infected Bacteria 



Virus lysate 


Analyses 


Initial total 


Total lysate 


Low speed 
sediment 


High speed 

supernatant 

fluid 


High speed 
sediment con- 
taining virus 






mg. 


mq. 


per cent 


per cent 


per cent 


T4A, 97 cc. 


P 




0.210 


31 


29 


40 




DNA 


0.213 


1.49 


25 


37 


38 


T4B, 105 " 


P 


0.141 


0.263 


19 


21 


60 




DNA 


0.220 


1.80 


17 


21 


62 


T2C, 96 " 


P 


0.161 


0.275 


27 


22 


51 




DNA 


0.275 


1.63 


29 


20 


51 


T2D, 92 " 


P 


0.148 


0.260 


33 


17 


50 




DNA 


0.221 


1.36 


35 


16 


49 



T4A 1.05, T4B 1.00, T2D 1.02. Thus the conditions described above yielded 
fractions with the characteristics of virus whose inorganic P contents were 
so low as to be undetectable. 

Infected cells may synthesize as much as 7 to 8 times as much DNA as 
was originally present. In many cases, at least 3 times as much DNA was 
isolated in the virus as was present in the bacteria at the onset of infection. 
It is considered probable that DNA formed after infection and not recov- 
ered in isolated virus was nevertheless originally part of the virus and was 
lost either by adsorption to debris or by becoming insoluble as a result of 
high speed centrifugation. No indication has been obtained in prepara- 
tions of either T2r+ or T4r"^ bacteriophage from F lysates that a phosphoryl- 
ated compound other than DNA is present in either virus. Furthermore, 
in contrast to the observations of Taylor (5), preparations of these viruses 
from nutrient broth lysates prepared from infected cells as described above 
also possessed DNA-P to total P ratios of 1.0 (6). Hence no evidence has 



118 



298 



BACTERIAL VIRUSES. II 



been obtained to confirm the report of Taylor that T2r"'" virus contains 
ribonucleic acid (5). It is considered possible that the method of isolating 
virus from 5 to 6 hour lysates described above assists in the removal in the 
low speed sediment of bacterial components containing RNA. Since pro- 
teolysis in r+ systems is relatively weak, the prolonged period used by 
Hook et al. (7) in the preparation of lysates may have assisted the degra- 
dation of bacterial debris to a size which would not sediment at a low speed 
but would at high speeds. Proteolysis in r"*" systems in F medium seems 
even less pronounced than in broth. 

Synthesis of Virus in Labeled Host Cells — Small inocula of bacteria were 
grown to 2 X 10* per cc. in 125 ahquots of F medium containing 0.02 to 0.05 
millicurie of P^- in 13.7 mg. of inorganic phosphate. More than 99 per cent 

Table II 
Radioactivity of Virus Isolated after Synthesis in P^^-Labeled Cells in Media Free of P^^ 



Virus isolated 


DNA-P 
Total P 
in virus 


Bacteria 


Virus 


Relative radioac- 
tivity, 
virus P 
host P 






counts per 10 y P per 
min. 


counts per 10 y P per 
min. 


per cent 


T2r+* 


1.00 


222 


33 


14 


T^r+t 


1.00 


293 


53.5 


18 


T4r+* 


1.01 


.37.8 


5.9 


16 


T4r+t 


1.06 


293 


53.3 


18 



*Virus to cell ratio 3.0. 
fVirus to cell ration 5.0. 

of the P of the bacteria was derived from the P of the medium. The cells 
were washed twice with 0.85 per cent NaCl. The second washing con- 
tained less than 0.1 per cent of the radioactivity of the original medium, as 
determined on a Geiger-Miiller counter, kindly loaned by Dr. H. D. Brun- 
ner of the Department of Pharmacology of the University of Pennsylvania. 

The radioactive cells were resuspended in 131 cc. of F medium containing 
2.5 X 10~^ M tryptophan. Purified T2r+ or T4r+ virus was added and the 
infected cultures were incubated for 5 hours and regularly assayed for virus 
durmg this interval. The lysates were stored at 4° overnight and virus was 
isolated as previously described. 

The DNA contents of the initial cultures, final lysates, and isolated virus 
were determined. The radioactivity of the virus was compared with the 
radioactivity of the miinfected host cell per unit weight of P. Samples of 
known P content were digested in 1.2 cc. of 60 per cent perchloric acid and 
diluted to 10 cc. after the solutions were neutrahzed to pH 4. The counting 



119 



S. S. COHEN 



299 



chamber was filled by perfusion and held identical aliquots of slightly more 
than 3 cc. The background was determined between samples and yielded 
8 to 11 counts per minute. This was subtracted from the count on the 
sample. Samples and the time of counting were adjusted to yield 100 to 
1000 counts. Comparisons of samples were made at similar total counts. 
In Table II are presented data on four experiments of this type, in which 
the increments in DNA in the infected bacteria and the yields of isolated 
virus were comparable to the data presented previously. The radioactivity 
of the virus isolated after synthesis under these conditions was far lower 
than the radioactivity of the host P. It has therefore been concluded that 
most of the P organized into virus is derived from the medium after in- 
fection. 

Table III 
Radioactivity of Virus Synthesized in Unlabeled Cells in Media Containing P^^ 



Virus isolated 


DNA-P 
Total P 
in virus 


F medium 


Virus 


Relative radioac- 
tivity 
virus P 
medium P 






counts per 10 y P per 
min. 


counts per 10 y P per 
min. 


per cent 


T2r+* 


1.04 


443 


315 


71 


T4r+t 


1.00 


840 


633 


75 


T4r+* 




422 


291 


69 



*Virus to cell ratio 5.0. 
fVirus to cell ratio 3.0. 

Synthesis of Virus in Unlabeled Host Cells in Media Containing Radio- 
active Phosphate — Bacteria were grown in the absence of radioactive P. 
Virus was added immediately after the addition of 0.02 to 0.05 millicurie of 
inorganic phosphate and tryptophan for the adsorption of T4. The cul- 
tures were incubated, and the DNA increments, rates of virus liberation, 
and isolated virus were studied as described previously. 

In Table III are presented some data on experiments of this type. The 
radioactivity of the virus isolated after synthesis under these conditions was 
70 to 75 per cent of that of the inorganic phosphate of the medium. These 
data confirm the previous conclusion that most of the P organized into 
virus is derived from the medium after infection. 

Nucleic Acid Turnover after Infection — Since the P^^ added to the F me- 
dium was assimilated in DNA after infection, the distribution of the radio- 
activity in the infected cell was studied. In a typical experiment bacteria 
were grown to 2 X 10^ per cc. in 1 liter of F medium. The cells were centri- 
fuged and resuspended in 100 cc. of F medium. To the culture were added 



120 



300 BACTERIAL VIRUSES. II 

4.5 cc. of T2r+ at 2.2 X 10^ ^ per cc. and 1.5 mg. of P in inorganic phosphate 
containing about 0.05 milhcurie of P^\ The infected culture was aerated 
vigorously at 37° for 1 hour. At zero time and at 1 hour, aliquots were 
analyzed for protein-bound P, DNA, and radioactivity. 

The counts per minute of the medium were 93 per microgram of P. In 
1 hour, the bacterial protein-bound P increased from 0.0109 mg. per cc. to 
0.0163 or 5.4 y per cc. Newly assimilated P had a radioactivity of 98 
counts per microgram. The DNA had increased from 24 to 72 y per cc, 
or the ratio of newly formed DNA to original DNA was 2:1. 

Following the removal of aliquots after 1 hour, the entire culture was pre- 
cipitated with 5 per cent trichloroacetic acid (TCA) and the sediment was 
washed twice in the centrifuge with TCA. The supernatant fluids were 
pooled. The sediment was washed twice with 95 per cent alcohol and twice 
with ether. These supernatant fluids were combined with lipide extracts 
prepared by three extractions of this sediment with boiling alcohol-ether 
(1:1). The pooled lipide extracts were taken to dryness in vacuo and dis- 
solved in 10 cc. of CHCI3. 

The dried lipide-free sediment was fractionated according to Schmidt and 
Thannhauser (8). It was incubated with 2 cc. of n KOH at 37° for 20 hours 
and chilled. To the solution containing a very slight flocculent precipitate 
were added 0.4 cc. of 6 n HCl and 2 cc. of 5 per cent TCA. After 30 minutes 
at 0°, the sediment, containing DNA, was removed by centrifugation and 
washed with several cc. of water. The DNA fraction was dissolved in 0.1 
N NaOH. The combined supernatant fluids were precipitated according 
to Delory (9) ; essentially no P was found in this sediment and this fraction 
was discarded. The supernatant fluid contained ribose-3-phosphate nu- 
cleotides and comprised the RNA fraction. 

Both the DNA and RNA fractions prepared above were analyzed for 
DNA. The former contained 6.9 mg. and the latter 0.6 mg. of DNA, or a 
total of 7.5 mg. of DNA was recovered. The total DNA content of the 
culture was 7.4 mg. Radioactivity measurements of the RNA fraction 
were corrected for the radioactivity of the DNA it contained. This com- 
prised 10 per cent of the P of this fraction. The radioactivities of the 
nucleic acid fractions are summarized in Table IV. 

One-half of the P in the RNA fraction consisted of ribose-3-phosphate 
nucleotides reactive in the Bial reaction (10). Pyrimidine nucleotides de- 
rived from RNA and comprising one-half of the total P would not be 
expected to react. Thus the RNA fraction with the exception of a smaU 
amount of DNA contained ribose nucleotides of the expected character- 
istics. 

It may be seen from Table IV that the newly synthesized DNA contained 
P of essentially similar radioactivity to that in the medium, while the RNA 



121 



S. S. COHEN 



301 



fraction contained essentially no radioactive P. It may be concluded from 
this experiment and others which yielded the same result that not more 
than 2 per cent of the RNA could have been imdergoing synthesis involving 
P of the medium. It is highly improbable that the turnover of this much 
RNA, if any synthesis actually occurred, could have produced the synthesis 
of the DNA found. Therefore it appears that RNA nucleotides are not 
precursors to DNA synthesis in this system. 

Since special precautions were not taken to purify the nucleotides of the 
RNA fraction before estimating the radioactivity, it is possible that even 
the low radioactivity found (2 per cent) was due to some contaminant, and 
was not truly significant. In this experiment the lipide P was found to 
have a radioactivity of about 26 per cent of the P of the medium. Repre- 
cipitation of phospholipide in the presence of carrier sphingomyelin derived 

Table IV 
Inclusion of Radioactive P into Nucleic Acids of Escherichia coli B Infected with T'2r+ 

Bacteriophage 



Nucleic acid 


Total Protein-bound 
P 


Lipide-free bacterial 

residue 


Counts per 7 P per 
min. 


Corrected counts per 
7 P per min. 


DNA 
RNA 


per cent 

48 
36.5 


per cent 

13.8 
10.5 


60 

8.8 


90 
2 



DNA, counts corrected for presence of original bacterial DNA; RNA, counts cor- 
rected for presence of traces of DNA in RNA fraction. 

from beef lung readily reduced the radioactivity sufficiently to suggest the 
inactivity of the phospholipide fraction. 

Fractionation of Non-Protein P — It was not found possible to isolate or- 
ganic phosphate from the TCA supernatant fluids in this experiment, owing 
to the large excess of inorganic P of the medium. In another experiment, 
the bacteria were first sedimented to remove the inorganic P of the medium 
before fractionation. 10 cc. of TCA extract contained 66 7 of organic P 
and 183 7 of inorganic P. To this solution, adjusted to pH 8.3 with 
NH4OH, were added 1 cc. of 2.5 per cent CaCl2 and 2 cc. of a 0.5 per cent 
suspension of MgCOs. After being chilled, the mixture was centrifuged and 
the sediment was discarded. The supernatant fluid was precipitated with 4 
volumes of alcohol and stored overnight in the ice box. The alcohol precipi- 
tate was placed in 3 cc. of water. It contained 44 7 of organic P, of which 
ribose-5-phosphate in some form comprised 75 per cent of the total organic 
P of the fraction (10). The water-soluble fraction of the alcohol precipi- 
tate contained 16.5 7 of pentose P in 25 7 of organic P. The radioactivity 
per microgram of P of this fraction was 37 per cent of that of the inorganic 



122 



302 BACTERIAL VIRUSES. II 

P of the medium. It is not known in what compounds the radioactivity 
resided. 

DISCUSSION 

It may be seen from the data presented in this paper that the hypotheses 
suggested in Paper I have been verified by means of the isotope technique. 
These are (1) that phosphorylated virus constituents are synthesized in the 
main from P assimilated from the medium after infection, and (2) that RNA 
after infection has a very low, if any, turnover rate, and is not a precursor 
of virus DNA. 

Nevertheless the radioactivity of the virus P was significantly different 
from that of the P of the medium. It has been observed that a very small 
amount of virus multiplication may occur in the absence of external P (11). 
Therefore it appears that some of the P of the host may be incorporated into 
virus. It is considered likely that this small amount of P is derived from 
the intracellular pool of inorganic P or low molecular weight organic P 
which can equilibrate with P assimilated after infection. This intracellular 
metabolic pool would be used for the synthesis of DNA after infection and 
possibly even before the P assimilated after infection. Thus in short term 
experiments involving r strains the percentage of host P appearing in virus 
may conceivably be greater than that observed in these long term experi- 
ments with r+ type virus. This remains to be tested. 

Many workers have found that protein and DNA syntheses are accom- 
panied by a vigorous RNA metabolism in a wide variety of tissues. In 
addition it has been reported that there is apparently a conversion of ribose 
nucleotides to desoxyribonucleotides in the early stages of cleavage of the 
fertilized sea-urchin egg (12). These data are ably reviewed by Brachet 
(13). In the system described in these papers, none of these phenomena are 
to be observed. It appears possible to study the precursors of DNA in this 
system uncomplicated by the metabolism of RNA. 

I am indebted to Dr. Samuel Gurin of this University and Dr. M. Kamen 
of Washington University for their advice and assistance in the course of 
these studies. I wish to thank Miss Catherine Fowler for her technical 
assistance. 

SUMMARY 

The preparation and properties of bacterial lysates after multiple infec- 
tion have been described. As a result certain isotope experiments requiring 
viral isolation could be performed. The distribution of the bacterial and 
viral components in these lysates has been studied. It has been found pos- 
sible to recover in a purified concentrate of virus much of the P assimilated 



123 



S. S. COHEN 303 

and the DNA synthesized after infection. The newly synthesized DNA 
greatly exceeded the starting DNA of the bacterial culture. 

With P^2 in the labeled host or labeled medium, the DNA of the virus 
has been found to be built in the main from the inorganic P of the medium. 
Ribonucleic acid was not a precursor in the synthesis of this DNA and was 
essentially inert in infected cells. 

BIBLIOGRAPHY 

1. Cohen, S. S., J. Biol. Chem., 174, 281 (1948). 

2. Anderson, T. F., J. Cell, and Comp. Physiol, 25, 17 (1945). 

3. Cohen, S. S., and Anderson, T. F., /. Exp. Med., 84, 511 (1946). 

4. Cohen, S. S., J. Biol. Chem., 168, 511 (1947). 

5. Taylor, A. R., /. Biol. Chem., 165, 271 (1946). 

6. Cohen, S. S., Cold Spring Harbor symposia on quantitative biology. Cold 

Spring Harbor, 12, 35 (1947). 

7. Hook, A. E., Beard, D., Tavlor, A. R., Sharp, D. G., and Beard, J. W., J. Biol. 

Chem., 165, 241 (1946). 

8. Schmidt, G., and Thannhauser, S. J., J. Biol. Chem., 161, 83 (1945). 

9. Delory, G. E., Biochem. J., 32, 1161 (1938). 

10. Albaum, H. G., and Umbreit, W. W., /. Biol. Chem., 167, 369 (1947). 

11. Fowler, C. B., and Cohen, S. S., J. Exp. Med., in press. 

12. Brachet, J., Arch. Biol, 48, 520 (1937). 

13. Brachet, J., Embryologie chimique, Li^ge (1944). 



124 



ENZYMATIC SYNTHESIS OF DEOXYRIBONUCLEIC ACID. VI. 

INFLUENCE OF BACTERIOPHAGE T2 ON THE SYNTHETIC 

PATHWAY IN HOST CELLS* 

By Arthur Kornberg, Steven B. Zimmerman,^ S. R. Kornberg, and 

John Josse* 

DEPARTMENT OF MICROBIOLOGY, WASHINGTON UNIVERSITY SCHOOL OF MEDICINE, ST. LOUIS 

Communicated April 20, 1959 

Information now available about DNA^ synthesis by Escherichia coli enzymes^"* 
has encouraged an inquiry into the biochemical basis for the observation that a 
phage-infected E. coli cell ceases to produce its own DNA and makes instead the 
DNA characteristic of the infecting phage. ^ This general problem poses several 
rather specific questions which may be summarized as follows : 

1. T2, T4, and T6 DNA differ from the DNA of E. coli, as well as from that 
of other sources which have been examined, in containing hydroxymethylcytosine 
(HMC) but no cytosine.* Flaks and Cohen^ have already shown that within 
several minutes after infection by phage T2, T4, or T6, a new enzyme which hy- 
droxy methylates deoxycytidine 5 '-phosphate is produced. Is there an enzyme 
for converting the resulting dHMC-5-P to the triphosphate level in order to provide 
a functional substrate for DNA synthesis? 

2. With respect to the exclusion of cytosine from the DNA of phage T2, T4, and 
T6, is there a mechanism in the infected cell for removal of deoxycytidine triphos- 
phate from the site of polymerase action? 

3. The DNA's of T2, T4, and T6 contain glucose linked to the hydroxymethyl 
groups of the HMC in characteristic ratios,*' ^' '" although it is clear that in T2 
and T6 some of the HMC groups contain no glucose.^ According to our present 
understanding of DNA synthesis,^ it is difficult to conceive how these constant 
ratios are achieved if the incorporation were to occur via glucosylated and non- 
glucosylated HMC nucleotides. Is there an alternative mechanism involving 
direct glucosylation of the DNA even though direct substitutions on intact DNA 
have been hitherto unknown? 

4. Following phage T2 infection there is a temporary halt followed by a resump- 
tion of DNA synthesis at about 5 times the rate shown by the uninfected cell. " 
However, measurements with extracts of infected cells, using standard substrates, 
revealed much diminished rather than the anticipated augmented levels of DNA- 
synthesizing activity.^- What are the altered conditions for assay of DNA syn- 
thesis in infected cell extracts which would elicit the high levels of activity ex- 
pected from the physiologic studies? 

We have explored these questions and have found that following infection with 
phage T2 several new enzymes appear. ^^ These are (1) an enzyme which phos- 
phorylates dHMC-5-P, leading to the synthesis of hydroxymethyldeoxycytidine 
triphosphate (dHMC-TP), (2) an enzyme which removes the terminal pyro- 
phosphate group from dCTP, and (3) an enzyme which transfers glucose from 
UDPG directly to the HMC in DNA. Measurements of DNA synthesis, using 
dHMC-TP in place of dCTP, revealed about a 12-fold increase in activity in 
extracts of infected cells over the levels observed in uninfected cell extracts. 

Reprinted by permission of the authors and the National Academy 

of Sciences from the Proceedings of the National Academy of 

Sciences, 45 (6), 772-785, June, 1959. 

125 



Vol. 45, 1959 BIOCHEMISTRY: KORNBERG ET AL. 773 

METHODS AND MATERIALS 

Preparation of Cell Extracts. — E. coli B was grown at 37° with vigorous aeration 
in M-9 medium^* modified to contain per liter: KH2PO4, 3 gm, Na2HP04, 6 gm, 
NH4CI, 1 gm, MgS04-7H20, 0.49 gm, glucose, 5.0 gm, FeS04-7H20, 0.5 mg, CaClz,' 
55 mg. Growth at a logarithmic rate continued to 4-5 X 10^ cells per ml with a 
generation time of about 50 min. 

Cell extracts were prepared in two ways : 

Method I: Cultures grown to 2 X 10* cells per ml were chilled, centrifuged, and 
the cells resuspended at 4 X 10^ cells per ml in cold growth medium. Five T2r+'^ 
or T5 per cell were added and after a 4-min adsorption period at 0°, the culture was 
diluted twenty-fold into fresh growth medium at 37° and aeration continued. The 
time of dilution was taken as "zero minutes." 50-ml aliquots were pipetted rapidly 
onto crushed ice at intervals. The cells were sedimented by centrifuging for 5 min 
at 10,000 X g, M'ere resuspended in 1 ml. of 0.5 M glycylglycine buffer, pH 7.0, 
containing 0.001 M glutathione, and were stored for 1 to 3 days at -15°. The 
cells were disrupted in a 10 kc Raytheon sonic oscillator. After removal of a small 
amount of debris by centrifugation, the extracts contained about 2 mg of protein per 
ml. 

Method II: Cultures were grown to 2 X 10» cells per ml, in the modified medium 
without CaCl2 and four T2r+ per cell added ("zero minutes"). 50-ml aUquots 
were pipetted rapidly onto crushed ice at intervals. The cells were sedimented and 
resuspended in 4 ml of 0.05 M glycylglycine buffer, pH 7.0, containing 0.001 M 
glutathione and disrupted as above. Extracts containing about 6 mg of protein 
per ml were obtained after centrifugation. 

All results refer to extracts prepared by Method I unless otherwise stated. While 
Method II was less effective for phage multiplication (see below), a description of 
this method is included since it provided an alternative and efficient technique for 
obtaining concentrated extracts in kinetic studies. The activities per mg protein 
for the several enzymes studied were found to be at levels similar to those obtained 
by Method I. 

Bacteriophage Determinations. — Bacteriophage was assayed by standard tech- 
niques. '^ Intracellular phage was measured after "lysis from without" essentially 
as described by Doermann.^^ The formation of infectious units in both T2r+ and 
T5 infected cells (Fig. 1) was found to proceed in normal fashion in cells infected 
as described in Method I. T2r+-infections produced by Method II yielded only 
about 2 phage per original cell at 25 min when measured after "lysis from without," 
although after clearing of the culture a yield of several hundred phage per original 
cell was obtained. 

Enzyme Assays and Preparations. — Phosphorylation of the deoxynucleoside 
monophosphates (kinase activities) was measured, as described before, ^ by using 
a 5'-P^Mabeled mononucleotide as substrate and assaying the amount of label 
which becomes resistant to the action of semen phosphatase.^* Formation of 
dHMC-5-P from dC-5-P (hydroxymethylase) was assayed according to Flaks and 
Cohen.7 The assay of DNA synthesis ("polymerase") was measured by the con- 
version of a C^^-labeled deoxynucleoside triphosphate into an acid-insoluble product.^ 
"Polymerase" fraction VII from uninfected E. coli was prepared as previously 
described.^ 

126 



774 



BIOCHEMISTRY: KORNBERG ET AL. 



Proc. N. a. S. 



Substrates. — Deoxynucleotides and samples of native and enzymatically syn- 
thesized DNA were prepared as in earlier studies. ^^ ^ dHMC-5-P was synthesized 
according to Flaks and Cohen^ using a 230-fold purified hydroxymethylase^^; C'"- 
dHMC-5-P was obtained by using C'-formaldehyde (Volk Radiochemical Co.) 
in the hydroxymethylation reaction, and P^2_jjj]y[Q_5_p ^y^s prepared by using 
p32_(jQ_5_p The Em value determined for the nucleotide in this preparation 
was 13.5 X 10^ at 284 m/x at pH 1. Since this value conflicts with that of 11.7 X 
10^ given by Flaks and Cohen,' it is regarded as provisional and requires further 




20 30 

TIME - minutes 

Fig. 1 .— Appearance of pha{;e in T2- or T5-iiifected cells. In- 
fected culture.s were prepared by Method I. Infectious units were 
measured after "lysis from without." 

checking. C^^-glucose 6-phosphate was prepared by hexokinase action on uni- 
formly labeled C^^-glucose (Isotope Specialties Co.). C^^-UDPG was prepared 
from C ^"-glucose 6-phosphate and uridine triphosphate by the action of phospho- 
glucomutase and UDPG pyrophosphorylase as outlined by Glaser and Brown. 2» 
Unlabeled UDPG was a product of the Sigma Chemical Company. 

RESULTS 

An Enzyme which Phosphorylates dHMC-5-P. — At about 4 min after infection 
of E. coli with phage T2, it was possible to detect in the extracts an enzyme w^hich 
catalyzes the phosphorylation by ATP of dHMC-5-P (Fig. 2 A). This reaction was 



127 



Vol. 45, 1959 



BIOCHEMISTRY: KORNBERG ET AL. 



775 



undetectable (<0.5 per cent of the maximal value after infection with T2) in normal 
cells or in extracts of cells infected with T5 (Fig. 2 A), a, phage without HMC in 
its DNA. The maximal level of phosphorylating ("kinase") activity for dHMC-5- 
P was of the same order as that of the kinases for the other deoxynucleotides in- 



1 




10 20 30 40 



5 I 10 20 30 40 
Tl ME-minutes 



10 20 30 40 



Fig. 2. — Deoxynucleotide-phosphorylating enzymes ("kinases") and hydroxymethylating 
enzyme levels before and after infection with phage T2 or T5. The arrow indicates the start 
of infection ("zero minutes," see Method I). Assays were as referred to in Methods. 



128 



776 



BIOCHEMISTRY: KORNBERG ET AL. 



Proc. N. a. S. 



corporated into viral DNA (see below) and was essentially similar in extracts pre- 
pared by either Method I or Method II. 

The product of the dHMC-5-P kinase action (with the presumed participation 
of nucleoside diphosphate kinase in the preparation) was shown to be the triphos- 
phate. Using a 20-fold purified kinase preparation/^ 25 /unioles of dHMC-TP 
were prepared and isolated by ion-exchange chromatography, a yield of 92 per cent 
based on the starting dHMC-5-P. Theoretical specific radioactivity values were 
found in the isolated dHMC-TP when P32- or C '^-labeled dHMC-5-P was the 




10 20 30 

TIME- minutes 

Fig. 3. — dCTPase levels before and after infection with phage 
T2 or T5. The arrow indicates the start of infection ("zero 
minutes," see Method I). The incubation mixtures (0.25 ml) 
contained : 6.0 m/imoles of dCTP labeled with P'^ in the terminal 
pyrophosphate group (10' cpm per m/umole), glycine buffer, pH 
9.2 (0.04 M), MgCl2 (0.008 M), 2-mercaptoethanol (0.01 M) and 
a quantity of extract containing about 0.3 ng of protein. After 
20 min of incubation at 37°, the reaction was terminated by add- 
ing 0.5 ml of 0.1 A/^ HCl, followed by 0.2 ml of a mixture (contain- 
ing per ml: 5 mg of crystaUine bovine serum albumin, 25 Mmoles 
of sodium pyrophosphate and 25 ^nioles of potassium phosphate, 
pH 7), and 0.1 ml of a Norit suspension (20 per cent packed 
volume). The Norit was removed by centrifugation and the 
Supernatant fluid was assayed for radioactivity. 

starting material. After concentration with Norit, analysis revealed a ratio of 
HMC: total P: acid-labile P of 1.0: 3.2: 2.2, using an E„, value of 13.5 X 10^ 
at 284 mju at pH 1 . 

Kinase Levels for the Other Deoxynucleotides and Levels of the Hydroxymethylating 
Enzyme. — It is noteworthy that after T2 infection, the kinase levels for dT-5-P 
and dG-5-P were increased approximately 20 and 45 times, respectively, while 
that for dA-5-P was essentially unaltered (Fig. 2 B, C, D). Bessman has made 
similar observations independently.^' As mentioned above, the kinase levels for 



129 



Vol. 45, 1959 BIOCHEMISTRY: KORNBERG ET AL. 777 

each of the four deoxyriucleotides incorporated into phage T2 DNA reach values of 
about the same magnitude. By contrast, only traces of dC-5-P kinase were 
detected (Fig. 2 E). Furthermore, extracts of T2-infected cells Avere found to 
inhibit the dC-5-P kinase activity of normal cell extracts when equal amounts of 
infected and non-infected extracts were mixed. As will be described below, this 
inhibitory effect is due to an enzyme (dCTPase) which splits dCTP. By use of 
fluoride (8 X 10 ~^ M), which inhibits dCTPase more than 98 per cent, but the dC- 
5-P kinase by 15 per cent or less, it was possible to show that there was actually 
little or no change in the dC-5-P kinase levels upon infection. Extracts prepared by 
Method II showed the same kinase patterns after T2 infection. 

Infection with phage T5, which contains cytosine rather than HMC, showed in- 
creased kinase activities for the four deoxynucleotides which are present in its 
DNA (Fig. 2 B, C, D, E). The 10-fold increase in dC-5-P kinase activity in the 
extracts of T5-infected cells may be contrasted with the absence of any increase in 
this activity in the extracts of T2-infected cells. 

Hydroxymcthylating activity was first detected at 4 min after T2 infection, and, 
as predicted from the results of Flaks and Cohen,'' was absent from T5-infected cell 
extracts (Fig. 2/^). 

An Enzyme which Destroys Deoxycytidine Triphosphate (dCTPase). — The in- 
hibitor of dC-5-P kinase that develops upon T2 infection has been identified as 
an enzyme which splits dCTP by removal of the terminal pyrophosphate group 
(Fig. 3). Extracts of uninfected cells have 1 per cent or less of the dCTPase 
activity observed in extracts of T2-infected cells; the level of dCTPase activity 
of To-infected cells was the same as in normal cells. The relative insensitivity to 
fluoride of the dCTPase activity in normal cells makes it doubtful that it represents 
the same enzyme found in the T2-infected cells. 

After partial purification of the dCTPase, ^^ which reduced the level of an inor- 
ganic pyrophosphatase to 2 per cent of the dCTPase, it was demonstrated that 
the complete splitting of 0.92 ^tmole of dCTP was accompanied by the appearance 
of 0.74 iumole of inorganic pyrophosphate identified as Norit-nonadsorbable, acid- 
labile P and 0.15 Mmole of orthophosphate. With the purified enzyme preparation, 
the rates of cleavage of cytidine triphosphate and ATP were less than 1 per cent of 
that of dCTP ; the splitting of dHMC-TP was 1-2 per cent of that of dCTP. The 
Km of dCTP for the enzyme is in the region of 10 ~^ M. In view of this high afl^nity 
for dCTP and of the high level of enzyme activity compared to the dC-5-P kinase 
in the extract (cf. Fig. 2 E and Fig. 3), it is reasonable to suppose that the dCTP 
concentrations in the infected cell are reduced to an extremely low level. 

An Enzyme which Glucosylates the HMC of DNA. — An enzyme which transfers 
glucose from UDPG to DNA containing HMC was observed in extracts of T2- 
infected but not T5-infected or normal cells (Fig. 4). With the partially purified 
enzyme'^ UDPG could not be replaced by glucose, glucose 1-phosphate or glucose 
6-phosphate (Table I) ; similar results were also obtained with the crude extracts. 
The reaction requires HMC-containing DNA, AAhich for these experiments was 
enzymatically synthesized from dHMC-TP, dATP, dGTP, dTTP, purified poly- 
merase, and primer DNA derived from any one of several sources (calf thymus, 
E. coli, phage T2, phage <t)X174^^). When DNA enzymatically synthesized with 
dCTP in place of dHMC-TP, or when the glucosylated DNA from phage T2 itself 

130 



778 



BIOCHEMISTRY: KORNBERG ET AL. 



Proc. N. a. S. 



were used, no glucose fixation in DNA was detectable (Table 1). dHMC-5-P 
and dHMC-TP failed to substitute for HMC-containing DNA as glucose acceptors. 
For example, in an incubation mixture containing these three types of HMC com- 
pounds with 30-fold purified enzyme, the HMC-DNA fixed 46 per cent of the glucose 
of the UDPG, while dHMC-5-P and dHMC-TP fixed none «0.5 per cent). 




10 20 30 

TIME - minutes 

Fig. 4. — DNA-glucosylating enzyme levels before and after 
infection with phage T2 or T5. The arrow indicates the start of 
infection ("zero minutes," see Method I). The incubation mix- 
tures (0.20 ml) contained: 10 mjumoles of UDPG labeled uni- 
formly with Ci'' in the glucose residue (2 X 10^ cpm per mMmole), 
Tris buffer, pH 7.5 (0.1 M), glutathione (0.02 M), DNA synthe- 
sized enzymatically (using thymus or phage T2 DNA as primer) 
and containing about 1 mjumole of HMC, and extract containing 
10-50 ixg of protein. After 15 min of incubation at 30°, the mix- 
ture was treated as in the "polymerase" assay of incorporation 
of a labeled deoxynucleotide into an acid-insoluble product (see 
Methods). 

In the presence of an excess of HMC-DNA, the glucose of UDPG is transferred 
completely to the DNA. With an excess of UDPG and enzyme, the fixation of 
glucose in DNA reaches a limiting value, which is a function of the HMC-DNA 
present (Fig. 5). The number of glucose residues fixed in this experiment was 
approximately 60 per cent of the number of HMC residues in the added DNA. 
At this point it may be premature to regard the glucosylation limit observed with 



131 



Vol. 45, 1959 



BIOCHEMISTRY: KORNBERG ET AL. 



779 



TABLE 1 
Specificity op^ the Enzyme which Glucosylates DNA 

Expt. 
No. Conditions 

1 Complete system (10 m/xmoles of C'^-UDPG) 

Add C»2-UDPG (10 m/imoles). 

Add C"'-glucose (200 m^moles) 

Add Ci2-glucose 1-P (200 mMmoles) 

Add Ci2-glucose 6-P (250 m/imoles) 

Replace UDPG with C'''-glucose (10 m/xmoles) 

Replace UDPG with C'^-glucose 6-P (12 m/imoles) 

2 Complete system (DNA containing 0.46 m/tmole of HMC) 

Replace HMC-DNA with cytosine-DNA (containing 0.75 

m/xmole cytosine) 
Replace HMC-DNA with T2 DNA (containing 2 m^moles 

HMC) 

The complete system had the composition and was treated as described in Figure 4, using a 
30-fold purified enzyme, 0.3 fig in E.xpt. 1 and 1.5 ng in Expt. 2. In each case the DNA was 
glucosylated to its limit (see Fig. 5). 

* The glucose fixed is calculated on the basis of the specific radioactivity of the UDPG after dilution of the C" 
sample with Ci'-UDPG. 



Glucose fixed in 


DNA, 


m/imoles 





69 





65* 





53 





59 





59 


0.00 





00 





28 





00 


0.00 



O R _ 












C..0 








/ 




^ o r\ 








/ 




o20 

2 










1^ 

1 1 i\ 












z 

Q 






•/ 






c 












^ I.U 

0) 
X 












U. 




/• 








o n R 




1 
°-5aq enzyme protein 


3 






• = IOag enzynne protein 


0.0- 













1.0 2.0 3.0 4.0 

HMC-DNA -m/^, Moles of HMC 

Fig. 5. —Limit of glucose fixation in DNA as a function of the amount 
of DNA added. The experimental details were as in Fig. 4, using the 
partially purified enzyme. The HMC-DNA was prepared with thy- 
mus DNA as primer. 

a given sample of HMC-DNA as distinctive for the type of primer used in the 
enzymatic synthesis of the HMC-DNA. Further studies are required to determine 
how conditions of enzymatic polymerization, as well as the isolation of the DNA, 
may influence the glucose/HMC ratios obtained. 



132 



780 



BIOCHEMISTRY: KORNBERG ET AL. 



Proc. N. a. S. 



The C '"-glucose fixed in DNA was rendered acid-soluble by crystalline pan- 
creatic deoxyribonuclease. When 52 per cent of the nucleotides were no longer 
acid-precipitable, 88 per cent of the glucose had been made acid-soluble. When the 
C'-glucosylated HMC-DNA was digested to completion under conditions which 
Lehman has found^^ to yield a quantitative conversion of phage T2 DNA to 5'- 
mononucleotides, over 95 per cent of the radioactivity was found in the HMC 
deoxynucleotide fractions of the ion-exchange chromatogram. 

Increase Rate of DNA Synthesis upon Infection.— DNA synthesis, measured by 
the standard assay, ^ but with dHMC-TP instead of dCTP, is increased 12-fold in 
extracts prepared 19 min after infection with T2 (Fig. 6). Little or no DNA syn- 
thesis can be measured in these extracts when dCTP replaces dHMCTP. However 



o 

la 
o. 

d« 

E 



16 



12 



E 



0- 









^ 




- 






/ 








/ 










/ 








i 


J 


















-5 ^ 


, 1 

l( 


D 2 


1 




Fig. 6 — DNA "polymerase" levels before 
and after infection with phage T2 with 
dHMC-TP as substrate in place of dCTP. 
The arrow indicates the start of infection 
("zero minutes," see Method I). The incu- 
bation mixtures (0.30 ml) contained: 10 
niMmoles each of dGTP, dATP, dTTP and 
dHxMC-TP (C'S 1 X 103 cpm per m/imole), 
Tris buffer, pH 7.5 (0.07 M), MgCU (0.007 
M), 2-mercaptoethanol (0.001 M), 0.04 ml. 
of "heated DNA" and sonic extract con- 
taining 10-50 fig of protein. The "heated 
DNA" was prepared by heating at 100° 
for 5 min, at pH 9.2, an extract of T2- 
infected cells; this preparation could be 
purified without loss of activity by treat- 
ment with ribonuclease, Norit, dialysis and 
precipitation with acid; it was inactivated 
by treatment with deoxyribonuclease. Fur- 
ther details of procedure were as referred 
to in Methods. 



TlME-minutes 

using 0.005 M F^ to inhibit dCTPase, DNA synthesis was elevated to levels near 
those observed in assays with dHMC-TP. 

It should be emphasized that these measurements of rates of DNA synthesis 
with dHMC-TP were made with heated DNA as primer; when unheated DNA was 
used, there was no demonstrable increase in rate in the infected cell extracts. 
Heated DNA of phage T2 or calf thymus origin served as well as that used in Figure 
6. The basis for this requirement for heated DNA with infected cell extracts 
requires further investigation. 

DISCUSSION 

In addition to the interest inherent in understanding the nature of viral infection 
of a cell, studies of the pathway of DNA synthesis in infected cells provide a means 
of testing and expanding our conceptions about the mechanism of DNA replication 
in normal cells. 



133 



Vol. 45, 1959 BIOCHEMISTRY: KORNBERG ET AL. 781 

Control of DNA Synthesis at the Level of Deoxy nucleotide Phosphorylation. — 
It appears from studies with various analogues of the naturally occurring purines 
and pyrimidines^ that an analogue in the form of a deoxynucleoside triphosphate 
is incorporated into DNA when its structure has the hydrogen-bonding properties 
of the base it replaces. For example, uracil and 5-bromouracil are both effective 
substitutes for thymine. Yet, uracil is never found in DNA^* while bromouracil, 
when supplied to the cell, readily replaces thymine in the DNA.^^ The reason ap- 
pears to be the absence of any mechanism for phosphorylating deoxyuridylate as 
contrasted with the availability of an enzyme for phosphorylating 5-bromodeoxyuri- 
dylate or 5-methyldeoxyuridylate (thymidylate).^ 

The present studies of phage-infected cells provide additional examples of control 
at the phosphorylation level. Earlier observations with 5-methyldeoxycytidylate3 
had suggested the lack of an enzyme for phosphorylation of a 5-substituted cytosine 
deoxynucleotide, and posed the problem of how 5-hydroxymethylcytosine deoxy- 
nucleotide becomes a substrate for phage T2 DNA synthesis. This problem seems 
to be solved by the development of a new enzyme after phage T2 infection of the 
cell. According to our studies, and those of Flaks and Cohen,^ the synthesis of a 
compound novel for the cell is carried out by a new enzyme, the formation of which 
is presumably induced by the phage DNA. Another example of control at the 
phosphorylation level is provided by the device which appears to be responsible for 
the absence of cytosine in T2 DNA. For lack of a mechanism to eliminate the 
dCTP-synthesizing enzyme system, the evolution of an enzyme to destroy the 
dCTP provides an effective alternative. 

Direct Suhstitvtion on DA^^ .—Genetic studies indicate that the glucose contents 
of DNA of phages T2, T4, and T6, and strains derived from them by recombination, 
are a fixed property resembling other phenotypic characters of these phages. ^^ ^° 
It appears plausible, therefore, that the precise arrangement of these glucosylated 
HMC residues in the DNA may be a source of genetic information, and insight into 
the nature of the replication of these types of phage DNA would clearly be of con- 
siderable value. 

The incorporation of a fixed proportion of non-, mono-, and polyglucosylated 
derivatives of dHMC-TP during the polymerization of the DNA chain is difficult 
to conceive because it is likely that these derivatives would all behave similarly in 
hydrogen-bonding to guanine. It becomes even more difficult to conceive the in- 
corporation of these various HMC residues in a precise sequence in DNA on the 
basis of selection of the triphosphate derivatives. The existence of an enzymatic 
system for direct glucosylation of DNA offers an alternative which seems to circum- 
vent these difficulties. At this stage, the information is still too fragmentary to 
determine whether this glucosylating enzyme, and perhaps an additional one for 
polyglucosylation, will be sufficient to explain the replication of various phage 
DNA's. It is apparent that further studies with T4- and T6-, as well as T2-in- 
fected cell enzyme systems are essential. 

Superficially analogous to the partial glucosylation of the HMC residues in phage 
DNA is the partial methylation of the cytosine residues in certain plant and animal 
DNA's, such as wheat germ and calf thymus. 2" In the light of our findings, it 
would be interesting to look for an enzymatic mechanism for direct methylation of 
DNA in these tissues. 

134 



782 BIOCHEMISTRY: KORNBERG ET AL. Proc. N. A. S. 

Kinetics of Enzyme Development and Enzyme Reactions. — The temporal pattern 
of development of all the enzymes studied in this report that were either "new," 
or the levels of which were significantly raised, was similar. The first traces of 
change were apparent at four minutes after T2 infection and about ten minutes after 
T5 infection. While the precision of measurement of these time intervals is not 
great, it is clear that there is a time lag before significant levels of these enzyme 
activities appear. Several groups of investigators^^' "• ^^ have shown that 5- 
methyltryptophan and chloramphenicol administered to cells at levels sufficient 
to inhibit protein synthesis also blocked DNA synthesis when given before or up 
to about 5 min after infection. When given 10 min or later following T2 infection, 
the inhibition of protein synthesis had only a small or even no effect on the rate of 
DNA synthesis. It seems reasonable to consider that the time lag we have ob- 
served may be related to the chloramphenicol-sensitive interval required for 
development of the enzymatic machinery for DNA synthesis, as well as other 
protein components vital to virus information. There have been indications of 
non-phage protein synthesis immediately upon infection ^^ and it remains to be 
determined what fraction of this protein can be identified with the enzymes de- 
scribed here. 

The multiplicity of enzyme changes described in this study along with the 
findings on the hydroxymethylating^ and deo.xyuridylate-methylating^" enzymes, 
are all directly related to DNA synthesis. Enzyme measurements related to 
other metabolic pathways in infected cells have as yet disclosed few significant 
changes." It would be surprising if further exploration of the phage-infected 
cell failed to reveal additional examples of new or augmented enzyme activities 
related to the requirements imposed by rapid phage synthesis. 

In view of the pitfalls inherent in assaying the level of an enzyme activity in a 
cell extract, let alone in the cell iteelf, a detailed evaluation of the various enzyme 
values in terms of virus DNA synthesis does not seem warranted. However, it 
is interesting to note that the rates of the kinases and the glucosylating enzyme 
are all greater than that found for the DNA polymerizing activity and, further, 
that the increase in the latter activity over levels found in uninfected cell extracts 
is about the same as the increase in DNA synthesis in whole cells following T2 
infection. Also it is remarkable that the dA-5-P kinase activity (very likely 
identical to adenylate kinase) is about 10 times that of the other deoxynucleotide 
kinases in the uninfected cell extract and does not change upon T2 infection, 
while after infection the other kinases and hydroxy methylase reach levels com- 
parable to the dA-5-P kinase. Finally, it is noteworthy that the dC-5-P kinase 
activity does not increase upon T2 infection but remains at the relatively low 
level observed in the uninfected cell, whereas it increases about 10-fold upon T5 
infection. 

References to increases in level of a preexisting enzyme carry no implication 
that the additional enzyme activity is identical to the old or even that more 
enzyme has been synthesized "de novo." To resolve this important point, it will 
be necessary to characterize isolated preparations of the normal- and infected- 
cell enzymes and to establish by tracer techniques that the enzymes developed after 
infection have, like induced enzymes,^- been synthesized from the amino acid pool. 

Control of DNA Synthesis by the Nature of the DNA Primer. — Perhaps the most 

135 



Vol. 45, 1959 BIOCHEMISTRY: KORNBERG ET AL. 783 

interesting question regarding T2 infection of E. coli is how the T2 DNA seems 
to preempt from the host DNA the primer function for the polymerase system. 
The present studies can explain why DNA synthesized by the phage T2-E. coli 
system contains hydroxymethylcytosine rather than cytosine, but not why the 
base composition is that of phage rather than that of host DNA. An observation 
reported in this paper which may bear on this question is the dependence of the 
augmented polymerase activity of the infected cell extracts on a physically altered 
DNA primer. When "native" calf thymus or phage T2 DNA was used, the 
polymerase values of infected and normal cell extracts were similar. However, 
with DNA samples heated at pH 9, the values for dHMC-TP incorporation with 
infected cell extracts were increased 12-fold, while the normal cell values were 
unaffected or even depressed. It is not clear as yet whether infection has resulted 
in formation of a new type of polymerase or in an increase of the normal poly- 
merase and new factors which modify it. Further studies to clarify this question 
may lead to a better insight into how the injected phage DNA proves to be the 
chosen primer for this system. 

SUMMARY 

1. Extracts of E. coli infected with bacteriophage T2 have been shown to 
contain three enzymes which are undetectable in extracts of uninfected or in T5- 
infected cells. These are : (a) an enzyme which phosphorylates hydroxymethylde- 
oxycytidine 5-phosphate, leading to the synthesis of the triphosphate, (6) an enzyme 
which removes the terminal pyrophosphate group specifically from deoxycytidine 
triphosphate, and (c) an enzyme which transfers glucose from uridine diphosphate 
glucose directly to the hydroxymethylcytosine of certain DNA's. 

2. These new enzymes can account for (a) the availability of the triphosphate 
of hydroxymethyldeoxyc3nDidine for the enzymatic synthesis of T2 DNA, (6) 
the absence of deoxycytidylate from T2 DNA, and (c) the presence of glucose on a 
fixed fraction of the hydroxymethylcytosine residues in DNA. 

3. The DNA-polymerizing enzyme assayed with hydroxy methyldeoxycytidine 
triphosphate in place of deoxycytidine triphosphate reveals about a 12-fold increase 
in activity after T2 infection. Increases, after T2 infection, in the levels of thymine 
and guanine deoxynucleotide phosphorylating enzymes (by 20-45 fold) bring 
their activities up to the level of the adenine deoxynucleotide phosphorylating 
enzyme which is unchanged ; the level of the cytosine deoxynucleotide phosphorylat- 
ing enzyme remains at a low level. 

4. After T5 infection, levels of the thymine, guanine, and cytosine deoxynucleo- 
tide phosphorylating enzymes increase by 10-40 fold, bringing their activities up 
to the level of the adenine deoxynucleotide phosphorylating enzyme, which in- 
creases about 2-fold. 

5. The new enzymes and the increases in level of the enzymes occurring in 
normal cells are first detectable about 4 min after infection with phage T2 and 
about 10 min after infection with phage T5. These results are consistent with 
previously published studies which have indicated with the use of inhibitors of 
protein synthesis that viral DNA synthesis requires a preliminary period of pro- 
tein synthesis. 

136 



784 BIOCHEMISTRY: KORNBERG ET AL. Proc. N. A. S. 

* This investigation was supported by research grants from the National Institutes of Health 
of the Public Health Service and the National Science Foundation. 

t National Science Foundation Predoctoral Fellow. 

X Fellow of the National Foundation. 

' The abbreviations used in this report are: cpm, counts per minute; ATP, adenosine tri- 
phosphate; dA-5-P, deoxyadenosine 5 '-phosphate; dC-5-P, deoxycytidine 5 '-phosphate; dG-5-P, 
deoxyguanosine 5'-phosphate; dHMC-5-P, hydroxymethyldeoxycytidine 5'-phosphate; dT-5-P, 
deoxy thymidine 5 '-phosphate; dATP, deoxyadenosine triphosphate; dCTP, deoxycytidine tri- 
phosphate; dGTP, deoxyguanosine triphosphate; dHMC-TP, hydroxymethyldeoxycytidine tri- 
phosphate; dTTP, deoxythymidine triphosphate; DNA, deoxyribonucleic acid; HMC, 5- 
hydroxymethylcytosine; UDPG, uridine diphosphate glucose; Tris, tris-(hydroxymethyl)amino- 
methane; KPO4, potassium phosphate buffer; PRPP, 5-phosphoryl a-D-ribofuranose 1-pyro- 
phosphate; E^, molar extinction coefficient. 

^ Lehman, I. R., M. J. Bessman, E. S. Simms, and A. Kornberg, /. Biol. Chem., 233, 163 
(1958). 

^ Bessman, M. J., I. R. Lehman, J. Adler, S. B. Zimmerman, E. S. Simms, and A. Kornberg, 
these Proceedings, 44, 633 (1958). 

^ Bessman, M. J., I. R. Lehman, E. S. Simms, and A. Kornberg, J. Biol. Chem., 233, 171 ( 1958); 
Adler, J., 1. R. Lehman, M. J. Bessman, E. S. Simms, and A. Kornberg, these Proceedings, 44, 
641 (1958); Lehman, L R., S. B. Zimmerman, J. Adler, M. J. Bessman, E. S. Simms, and A. 
Kornberg, these Proceedings, 44, 1191 (1958); Kornberg, A., Harvey Lectures, 53, 83 (1957- 
1958). 

•6 Hershey, A. D., J. Dixon, and M. Chase, /. Gen. Physiol, 36, 777 (1952-1953). 

6 Wyatt, G. R., and S. S. Cohen, Biochem. J., 55, 774 (1953). 

^ Flaks, J. G., and S. S. Cohen, Biochim. et Biophys. Acta, 25, 667 (1957); Federation Proc, 
17, 220(1958). 

8 Sinsheimer, R. L., Science, 120, 551 (1954); Volkin, E., /. Am. Chem. Soc, 76, 5892 (1954). 
' Sinsheimer, R. L., these Proceedings, 42, 502 (1956); Jesaitis, M. A., J. Exp. Med., 106, 
233 (1957); Federation Proc, 17, 250 (1958). 

1° Streisinger, G., and J. Weigle, these Proceedings, 42, 504 (1956). 

11 Cohen, S. S., J. Biol. Chem., 174, 281 (1948). 

>' Kornberg, A., L R. Lehman, and E. S. Simms, Federation Proc, 15, 291 (1956), and more 
recent unpublished observations. 

'^ An abstract of this work has appeared [Zimmerman, S. B., S. R. Kornberg, J. Josse, and A. 
Kornberg, Federation Proc, 18, 359 (1959)], as have abstracts regarding an enzyme which phos- 
phorylates dHMC-5-P (R. Somerville and G. R. Greenberg, Ibid., 327), dHMC-TP incorporation 
into DNA, and a suggested dCTP-degrading enzyme (J. F. Koerner and M. S. Smith, Ibid., 264) 
in T2-infected E. coli. 

i'* Anderson, E. H., these Proceedings, 32, 120 (1946). 

^^ We are indebted to Dr. Helen Van Vupakis for a generous gift of phage T2r "•". 

" Adams, M. H., in Methods in Medical Research, ed. J. H. Comroe, Jr. (Chicago: Yearbook 
Publishers, Inc., 1950), 2, 1. 

1' Doermann, A. H., J. Gen. Physiol, 35, 645 (1952). 

'* When fluoride was present in the kinase assays, the nucleotides were adsorbed to and eluted 
from Norit before phosphatase treatment in Stage II of the assay. 

" For large-scale enzyme preparations, 50-liter cultures at the late exponential phase (2 X 10" 
cells/ml), in modified M-9 medium (see Methods) but lacking CaCL, were treated with 3-4 T2r"'' 
per cell and 10 min later with 50 7 of chloramphenicol per ml. The culture was then chilled to 0° 
with ice over a 10-min period and harvested by centrifugation. The initial steps in the purifi- 
cation of the dHMC-5-P kinase, dCTPase, glucosylating enzyme, and hydroxy methylating 
enzyme were the same. Sonic extracts, prepared in 5 volumes of 0.05 M glycylglycine buffer, 
pH 7.0, were centrifuged and the supernatant fluid diluted with buffer to contain 10 mg of pro- 
tein per ml. Streptomycin sulfate (5 per cent solution), equal to 0.3 volume of the diluted ex- 
tract, was added; the supernatant fluid collected after centrifugation contained 3 to 4 mg of pro- 
tein per ml. It was adjusted to pH 8 with KOH and applied to a diethylaminoethylcellulose 
(Brown Co.) column [Peterson, E. A., and H. A. Sober, /. Am. Chem. Soc, 78, 751 (1956)], equi- 

137 



Vol. 45, 1959 BIOCHEMISTRY: KORNBERG ET AL 785 

librated with 0.02 M KPO4, pH 8.0, containing 0.01 M 2-mercapto(!thanol. Linear gradients 
were applied and the purifications obtained are expressed relative to t'.ie supernatant fluid of the 
sonic extract; all solutions contained 0.01 M 2-mercaptoethanol. Glucosylating enzyme: 
gradient of 0.08 to 0.32 M NaCl in 0.02 M KPO4, pH 8.0-30 X purification; dCTPase: gradient 
of 0.08 M to 0.32 M NaCl in 0.02 M KPO4, pH 8.0-24 X purification, rechromatographed with a 
gradient of 0.06-0.32 M KPO4, pH 6.5-85X purification; dHMC-5-P kinase: gradient of 0.06 
to 0.32 M KPO4, pH 6.5-20 X purification; hydroxymethylating enzyme: gradient of 0.3 M 
KPO4, pH 6.5 to 1.0 Af KPO4, pH 6. 1-230 X purification. Fractions containing 2-mercapto- 
ethanol were not frozen. Active fractions were concentrated by precipitation with solid (NH4)2- 
SO4. Further details of the purification procedures will be published elsewhere. 

2" Glaser, L., and D. Brown, these Proceedings, 41, 253 (1955); Glaser, L., J. Biol. Chem., 
232, 627 (1958). 

21 Personal communication from Dr. M. J. Bessman. 

22 We are indebted to Dr. R. L. Sinsheimer for the sample of <^X174 DNA. 
"^ Personal communication from Dr. I. R. Lehman. 

** Chargaff, E., in Nucleic Acids, ed. E. Chargaff and J. N. Davidson (New York: Academic 
Press, Inc., 1955), I, 307. 

« Dunn, D. B., and J. D. Smith, Nature, 174, 305 (1954); S. Zamenhof and G. GribofT, Nature, 
174,306(1954). 

26 Burton, K., Biochem. J., 61, 473 (1955). 

2' Tomizawa, J., and S. Sunakawa, /. Gen. Physiol., 39, 553 (1956). 

28 Hershey, A. D., and N. E. Melechen, Virology, 3, 207 (1957). 

" Watanabe, L, Biochim. et Biophys. Acta, 25, 665 (1957); Watanabe, I., and Y. Kiho, Proc. 
Int'l. Symp. Enzyme Chemistry (Tokyo-Kyoto, 1958), p. 418; Hershey, A. D., A. Garen, D. K. 
Eraser, and J. D. Hudis, Carnegie Institution of Washington Year Book, 53, 210 (1953-1954). 

^ Cohen, S. S., Abstracts of the American Chemical Societj^, Meeting, September, 1958, p. 22C. 

31 Pardee, A. B., and R. E. Kunkee, J. Biol. Chem., 199, 9 (1952). Dr. Fred Bergmann has 
found no changes, following T2 infection of E. coli B, in amino acid "activation" levels for valine, 
isoleucine, methionine, leucine, phenylalanine, tryptophan, and tyrosine. We have observed that 
levels of the following activities did not change ( ±20 per cent) within 15 min after E. coli B infec- 
tion with phage T2: PRPP synthetase, orotate -|- PRPP conversion to uridylate, inorganic pyro- 
phosphatase, adenine + PRPP conversion to adenylate, and adenylate kinase. 

32 Hogness, D. S., M. Cohn, and J. Monod, Biochim. et Biophys. Acta, 16, 99 (1955); Rotman, 
B., and S. Spiegelman, J. BactenoL, 68, 419 (1954). 



138 



THE FREQUENCY DISTRIBUTION OF SPONTANEOUS 

BACTERIOPHAGE MUTANTS AS EVIDENCE FOR 

THE EXPONENTIAL RATE OF PHAGE REPRODUCTION' 

S. E. Luria 

University of Illinois, Urbana, Illinois 

The phage geneticist is faced with the task of constructing a satisfactory model 
of phage reproduction, in the absence of direct morphological evidence similar to 
the one available to the macro-geneticist. Cell division, mitosis, meiosis, fertiliza- 
tion have a solid basis of morphological observation that the modern geneticist 
takes for granted. The virologist, on the other hand, begins where the cyto- 
geneticist ends; in a sense, he deals directly with the units of genetic material 
whose existence the macro-geneticists (including the bacterial geneticists) must 
infer. Here lies his weakness, since little is known of the performance of such 
units — and also his strength, since he can manipulate this subcytological world. 
He is not limited to dealing with integrated units of reproduction at the cellular 
level, but can control to a certain extent what goes into his cells. Because of this, 
virology's methods may lead more directly to solving the problem of the mode of 
repHcation of genetic material. 

Penetration of one phage particle into a susceptible bacterium leads to pro- 
duction of a large number of similar particles. The intervening steps are un- 
known. We conjecture a reorganization of the viral material, because of its 
nonrecoverability in infectious form early after penetration (Doermann, 1948). 
We conjecture an integration of the viral material into the cell machinery at the 
genetic level (Luria and Human, 1950), because cell syntheses are redirected 
toward the production of virus specific substances. The genetic complexities of 
bacteriophage tell us that the viral specificities to be replicated are multiple 
(Hershey, 1946b) and that the new virus may receive imprints from more than 
one viral ancestor within the same cell (Delbriick and Bailey, 1946; Hershey, 
1946b). There is evidence suggesting a discrete and regularly assorted nature of 
the material determinants of these specificities (Hershey and Rotman, 1948). 



lAided by grants from the American Cancer Society (upon recommendation of the Committee 
on Growth) and from the Research Board, University of Illinois. The competent assistance of 
Miss Martha R. Sheek is gratefully acknowledged. 

Reprinted by permission of the author and the Long Island 

Biological Association from Cold Spring Harbor Symposia on 

Quantitative Biology, 16, 463-470 (1951). 

139 



S. E. LURIA 

A theory proposed by this writer (Luria, 1947) assumed independent repUcation 
of these determinants (or groups of determinants) followed by their assembly 
into mature virus particles. This theory originally aimed at accounting for the 
reactivation of ultraviolet inactivated phage inside multiple-infected bacteria 
(Luria and Dulbecco, 1949) and was extended to account for some features of 
genetic recombination (Hershey and Rotman, 1949). The main ground for 
proposing the theory, namely, the belief that the reactivation resulted from 
genetic exchange, has been weakened by new evidence (Dulbecco, 1952), and the 
theory has as little left to support it as to disprove it. 

Yet, all this concerns the organization of virus material during reproduction, 
not the elementary process of replication. The latter cannot yet be attacked at 
the chemical level by any tool except speculation. It can be attacked on a 
limited front, however, by strictly genetic means. The experiments described 
in this paper were done to investigate the rate of replication of individual genetic 
determinants of the virus. They indicate that reproduction is exponential, each 
replica acting as a source of new replicas. 

Theory 

Phage mutations occur only in the intracellular state, presumably during 
replication. If a phage mutation occurs in a bacterium, that bacterium will 
liberate one or more mutants (assuming that no loss occurs intracellularly). 
Delbriick pointed out several years ago that the actual numbers of mutants 
would depend on the mode of phage replication — more specifically, on the mode 
of replications of the determinant or "gene" involved (see Luria, 1945b). We 
shall analyze a few possible mechanisms; other mechanisms may be proposed, 
but do not seem to lead to any simple picture. 

I. Exponential Reproduction 

One gene produces n genes, each of which in turn gives n genes, and so on. 
After r generations, the number will be n^ For n = 2 (duplication mechanism), 
the situation is analogous to bacterial reproduction. Let us assume this to be 
the case. N gene copies will derive from one gene by N-1 acts of replication. 
Let the last generation have the order number 0, the second last the order number 
1, the third last 2, and so on. Suppose a mutation occurs at generation k, either 
as an "error of replication" affecting one of the two products of duplication, or 
as a change in one gene during interphase. If the replication process is com- 
pletely synchronized (the consequences of nonsynchronization can, if necessary, 
be analyzed), the resulting clone of mutants will at generation consist of 2^ 
individuals. If the total number of individuals (at generation 0) in the popula- 
tion is N, at generation k there were N/2'' individuals. Assuming a constant 
probability m of mutation per individual, the number of mutations occurring 
at generation k is mN/2''. The relation between the number x of mutants in 
a clone and the frequency yx of such clones will then be obtained as follows: 

mN ^. mN /-1^ 

Yx = ^^; X = 2^; yx = ^1^ 

Z^ X 



140 



FREQUENCY OF SPONTANEOUS BACTERIOPHAGE MUTANTS 

Synchronization is expressed by the requirement that for x ^ 2^, y^ = 0. 

The frequency distribution (1) is, of course, identical to that of bacterial 
mutants in a series of similar populations (Luria and Delbriick, 1943; Lea and 
Coulson, 1949). If we limit our observation to one intracellular cycle of phage 
production, the maximum clone size will be the maximum "burst size." Each 
phage burst with 2^ particles represents a population. For m » 2~^, there will 
practically never be more than one mutation per bacterium; the frequency dis- 
tribution (1) will then be that of the mutant clones among a large number of 
bacterial yields, with limitations imposed by the inequality of burst size from 
cell to cell. 

II. Independent Successive Replications 
Let us suppose that each new replica of a gene is produced independently of 
the preceding one, for example, by a series of successive acts of replication 
controlled by the initial gene brought in by the infecting phage particle. 

(a) If one of the copies mutates (at the time of its formation or later) the 
probability of mutation in other copies produced in the same cell should not be 
affected. Assuming a uniform mutation probability m, the mutants will be 
distributed at random among phage bursts (Poisson distribution). 

(b) If the initial gene, the pattern, mutates while turning out replicas, we may 
assume that afterwards it produces only mutants. If the mutation rate is con- 
stant in successive replications, there will occur as many cases of mutation just 
before the production of the last viable gene copy as before the production of the 
second last, the third last, and so on. The mutants will be in clones, and the 
frequency of clones of different sizes will be uniform, at least up to the value of 
the minimum burst size. 

Experimental Work 

Previous data on the number of host range mutants in individual phage 
cultures (Luria, 1945a; Hershey, 1946a) did not allow the desired type of analy- 
sis, because the quantitative detection of the mutants was uncertain and because 
their low frequency made it necessary to look for them in mass cultures, where 
more than one cycle of intrabacterial growth of some mutants could take place. 

The experiments here reported consisted in making single infection of Escheri- 
chia coli B with phage T2L (Hershey, 1946a; for the experimental methods used, 
see Adams, 1950) and counting the mutants r oyw produced in phage bursts from 
individual bacteria. These mutant phenotypes (Hershey, 1946b) were chosen 
because they occur with suitable frequency, can be recognized and scored 
efficiently, and can be tested for genetic allelism or nonallelism (Hershey and 
Rotman, 1948). Previous extensive tests by Dulbecco (1949) and other tests 
made in the course of this work established the following technical points. 

1. Mutant plaques r and w can be detected and scored without difficulty after 
six to eight hours of incubation at 37°C on nutrient agar plates under standard 
conditions (plating in a 0.6% agar layer over 24 hour old 1% agar plates, with 
about 2 X 10* young bacteria per plate). Over 100 single plaque isolations and 
replatings confirmed the scorings made by plaque type. 



141 



S. E. LURIA 

2. A plaque of r type can be distinguished without difficulty from a "mottled 
plaque" stemming from a bacterium infected with a mixture of r and wild-type, 
even when r is in moderate excess. Thus, it is possible to distinguish a plaque 
originating from an r particle from one resulting from a mutation that occurs on 
the plate during the development of a wild type plaque, unless the r mutation 
occurred in the first bacterium infected after plating and produced within that 
bacterium a large majority of r mutants — an improbable occurrence, as the 
results of the present work will show. 

3. As for the w mutants, mixed plaques of w and wild type closely resemble the 
plaques of wild type, so that the chances of mistaking a mixed plaque for a pure 
w one are rather small. 

Several single burst experiments of T2L on B, with single mfection in Difco 
nutrient broth, gave average yields per bacterium around 60-100; the burst size 
distributions are shown in Table 1. The reasons for the lower average yield of 
T2L in Difco nutrient broth, as compared with the yields obtained several years 
ago in the same system (Delbriick and Luria, 1942), are unknown; they may 
have to do with changes in medium composition, in the bacterial host, or in the 
virus itself. 

Mass phage lysates often contain more phage particles than plaque-forming 
units; most of the particles can be caused to form plaques by treatment with 
distilled water or Zn++ before plating (Bertani, unpub.). Electron micrographic 
counts give, for carefully assayed lysates of T2L, ratios "particles/plaques" 
between 1.0 and 2.0 (Luria, Williams and Backus, 1951 and unpub.). The 
failure of some phage particles to form plaques is apparently a peculiarity of 
phage in mass lysates, probably due to a combination between phage and 
inhibitors of bacterial origin. Repeated attempts to reveal, by various treat- 
ments, any increase in the plaque count of dilute lysates similar to those used in 
the experiments here reported, constantly failed. It is likely that plaque counts 
reveal nearly 100 per cent of the phage particles plated. 

In our experiments, we plated the full content of tubes in which one or several 
bacteria had lysed. The loss of phage remaining in each tube after plating is of 
the order of five per cent or less. Thus, we feel that we recovered and examined 
practically the totahty of the active phage produced by the bacteria. 

Preliminary experiments indicated that mutants were present in abdut one 
burst out of 200. This made it possible to examine on each plate the phage yields 
from several infected bacteria, and yet to have almost never more than one 
mutant clone per plate. As many as 20 bursts per plate were examined in some 
experiments, particularly when only the r mutants were scored, since these are 
more easily recognized than the w mutants. 

Independently isolated r mutants are generally found to be nonallelic, yielding 
wild-type recombinants in mixed infection (Hershey and Rotman, 1948). To 
test for allelism among r mutants isolated in our experiments, stocks were pre- 
pared from individual mutant plaques and used, separately or m mixtures, to 
infect bacteria. The yields were examined for wild-type plaques. Similar tests 
with w mutants are technically more difficult and were therefore not attempted. 



142 



FREQUENCY OF SPONTANEOUS BACTERIOPHAGE MUTANTS 
TABLE 1. BURST SIZE DISTRIBUTION, T2L ON B, SINGLE INFECTION 



Date 



Number 

of 

plates 



Average 

number 

of bursts 

per 

plate 



Average 
yield 
per 

burst 



Median 



Range" 



12/7/50 


96 


0.93 


69 


101 


13-472 


12/8/50 


105 


0.89 


71 


84 


5-347 


4/10/51 


96 


0.73 


92 


106 


9-477 


4/23/51 


48 


0.85 


87 


111 


10-254 



*The maximum values are almost certainly too high, since they probably represent plates 
with two or more bursts. 

The Frequency Distribution of Mutants 

A total of 16 experiments, done between April 1949 and April 1951, yielded 
the data to be reported. In five experiments, only the r mutants were scored. 
The results are presented in Tables 2 and 3. A total of 90 plates with mutants r 
and 103 plates with mutants w were observed. The expected number of plates 
with both r and w mutants (calculated from the assumption of independent 
incidence of r mutations and w mutations, either in the same bacterium or in 
different bacteria whose bursts were pooled on one plate) was seven; six were 
found. 



TABLE 


2. SYNOPSIS OF 


EXPERIMENTS ON DISTRIBUTION 


OF 






r 


AND w MUTANTS 














Average 


Number 


Number 




Stock 


Number 


Number 


number of 


of plates 


of plates 


Date 


phage, 


of 


of 


bursts per 


with r 


with w 




number 


plates 


bursts 


plate 


mutants 


mutants 


4/26/49 


33 


144 


660 


4.6 


3 


10 


5/6/49 


1^ 


144 


660 


4.6 


7 


9 


3/9/50 


44 


142 


1120 


7.9 


12 


7 


3/14/50 


?? 


143 


1180 


8.25 


4 


15 


3/31/50 


19 


144 


2020 


14 


6 


18 


5/2/50 


19 


143 


250 


1.75 


2 


3 


5/19/50 


46 


144 


1060 


7.5 


9 


12 


5/26/50 


1^ 


144 


1800 


13 


15 


9 


12/18/50 


51 


605 


540 


0.89 


2 


1 


1/11/51 


" 


144 


800 


5.5 


7 


10 


2/12/51 


•>^ 


144 


820 


5.7 


4 


9 


3/1/51 


'» 


192 


1500 


7.9 


1 


* 


3/8/51 


" 


192 


2400 


17.5 


1 




3/20/51 


n 


180 


3400 


19 


7 




3/29/51 


n 


173 


4400 


25.5 


9 




4/10/51 


n 


96 

Totals 


70 


0.73 


1 
90 






22620 


103 



*Mutants w not scored. 



143 



S. E. LURIA 

Let us observe the frequency distributions of the numbers of mutants (Table 
3). The total number of mutants observed in all experiments (Table 3; total 
clone frequency) was 766, distributed among 2874 plates containing about 
1.8 X 10^ normal phages. The proportion of mutants is about 4 X 10"'*, and 
their average number per plate is about 2.5 X 10~^ If the mutants were dis- 
tributed at random, there would be about 550 plates with one mutant, about 90 
with two or three mutants, and only four with four mutants or more. There is 
no doubt that the distributions are not random, but clonal. 

As a test of the nature of the mutants appearing on the same plate, 11 pairs 
of r mutants were isolated from 11 plates, which contained between 2 and 59 
mutants. In all cases, the mutants in each pair proved allelic (probably iden- 
tical) ; no wild-type recombinant was observed among at least 1000 plaques of the 
yield from mixed infected bacteria. In 11 out of 12 crosses between mutants 
isolated from different plates we observed wild-type recombinants; the twelfth 
cross failed to show recombinants. It may have represented either a case of 
repeated occurrence of the same mutation or a case of two mutations with 
recombination frequency lower than 0.2 per cent, the lowest frequency detectable 
in our rather crude tests. These results, then confirm the clonal nature of the 
mutants produced within a given bacterium. We will now consider the clonal 
distributions. 

Inspection of Table 3 shows that the mutant distribution, though clonal, fails 
to fit the uniform frequency predicted for small clones of various sizes by the 
hypothesis of independent gene replication, with mutations occurring in the 
pattern. 

Let us turn next to the distribution predicted by the hypothesis of exponential 

duplication. 

Inspection of the data shows: (a) there are clones with two mutants; therefore, 
if exponential reproduction occurs, the elementary process is probably one of 
duplication (from 1 to 2) rather than triplication or quadruplication ; (b) there 
are clones with 3, 5, 6, 7, . . . in addition to clones with 1, 2, 4, 8, . . . mutants. 
Thus, exponential reproduction, if present, must be nonsynchronized, a conclu- 
sion also suggested by the well-known distribution of the total burst size. 

For a quantitative test of the hypothesis of exponential reduplication we shall 
use, instead of Equation (1), the following expression (accumulated distribution) 
suggested by Dr. Howard Levene : 

Y. = Sy. = Sf = sf = »NS ^.^0,-"^ (forN = 2k».). (2) 

Yx is the number of clones with x or more mutants. The product Yx X x is 
constant and a plot of log Y versus log x gives a straight Ime with slope - 1 ; the 
vertical intercept for x = 1 is the logarithm of the total number of mutant clones. 
This plot has the advantage that it is hardly affected by nonsynchronization. A 
clone with three mutants can be considered either as a clone that should have had 
two mutants and underwent one extra reduplication, or as a four-mutant clone 



144 



FREQUENCY OF SPONTANEOUS BACTERIOPHAGE MUTANTS 

that lagged one reduplication behind. In either case, the clone will contribute to 
the value Y2 and will be deducted from the value Y4. 

A correction should be made before comparing the results with the theoretical 
expectation. The stocks of phage used to infect bacteria contained small, known 

TABLE 3. THE DISTRIBUTION OF MUTANTS IN INDIVIDUAL PLATES 





Clone frequency, r mutants 


Clone frequency, w mutants 

Corrected for 
All plates full mutant 
clones 


Clone 

total, 

for fi 

( 


frequency, 


Clone 
size 


All plates 


Corrected for 

full mutant 

clones 


corrected 
ill mutant 
clones 


1 


47 


47 


46 


46 




93 


2 


9 


9 


18 


18 




27 


3 


11 


11 


11 


11 




22 


4 


2 


2 


4 


4 




6 


5 


2 


2 


1 


1 




3 


6 






7 


7 




7 


7 


2 


2 


2 


2 




4 


8 


1 


1 


1 


1 




2 


9 














10 


2 


2 


1 


1 




3 


11 






1 


1 




1 


12 


1 


1 








1 


13 














14 


1 


1 


1 


1 




2 


15 


1 


1 


1 


1 




2 


16 






1 


1 




1 


20 






2 


1 




1 


22 


2 


2 








2 


25 






1 








26 






1 


1 




1 


30 


1 


1 








1 


34 






1 


1 




1 


37 






1 








39 


1 












40 


1 




1 








41 


2 


1 








1 


47 


1 


1 


1 






1 


53 


1 












59 


1 


1 








1 


100 
Total 


1 

90 


85" 


103 


98 




183 



numbers of r and w mutants. Since we plated more than one burst per plate, 
some of the mutant clones observed may have stemmed from bacteria infected 
solely with one mutant particle. These clones ought to be eliminated (since they 
derive from mutant particles originated in a mass lysate, where different condi- 
tions obtain) but they cannot be recognized. The expected number of such "full 
mutant" clones was 10. Assuming that a "full mutant" clone would be at least 
as large as the minimum burst size, we eliminated from the experimeiits where 



145 



S. E. LURIA 

such clones were expected to be present an appropriate fraction of the largest 
clones (see Table 3). This correction is indeed a small one. 

Figure 1 compares the experimental distribution of clones with one or more 
mutants with the expected distribution from the hypothesis of exponential 
reduplication. The data corresponding to the "corrected" columns in Table 3 
are plotted for r mutants, for w mutants and for the two together. They fit well 
the expected relationship (Imear relation between log Y^ and log x with slope — 1) 
for low clone size, up to mutant clone sizes of the order of 10-15. Above that 
point, the frequency of mutant clones falls below the theoretical values. This 
behavior is precisely what we should expect. The linear relation between log y 
and log X — see Equation (1) — should only obtain for clones so small that they 
have equal chances to be formed in all bacteria. For clone sizes of the order of 
the burst size, a limitation is placed on the frequency with which these clones can 
be observed. In the curve log Yx versus log x this limitation will manifest itself 
as a downward concavity, which becomes appreciable aromid the value corre- 
sponding to the lowest class of frequent burst sizes (about 20 phages per bacteri- 
um in our experiments) and progressively more pronounced as the median burst 
size is approached. Of course, there cannot be any clone larger than the maximum 
burst size. An additional factor (suggested by Dr. S. Dancoff) that works in the 
same direction is nonsynchronization itself; in fact, this results in the existence, 
within each burst, of subclones that have origmated at the same generation but 
have different sizes, thus producing effects similar to those of the burst size 
differences. 

No closer analysis of the concave portion of the distribution frequency curve is 
feasible beyond these qualitative considerations, since the clones in this region 
are few and fluctuations affect the results strongly. Altogether, our results fit 
quite well the hypothesis that the genes responsible for the investigated pheno- 
types reproduce exponentially by successive reduplications. Let us now analyze 
some of the factors that might affect the experimental results. 

1. Failure to recognize mutants. This cause of error is difficult to assess; we 
believe the error to be very small. All plates were scored by the same observer 
after the optimum incubation period, and every plate that might have presented 
difficulties in scoring, because of crowding or of faulty layering, was discarded 
before exammation. All the doubtful plaques were picked and replated for the 
phenotype test. Any residual error from this source would probably result in 
underestimation of the frequency of clones with one mutant, since the finding of 
the first mutant on a plate might sharpen the alertness of the observer, thereby 
increasing the chances of detecting other mutants on the same plate; the classes 
of clones with more than one mutant might thus have been favored in our 
observations. 

2. A more definite and more easily evaluated source of error is the coincidence 
of more than one clone of a given mutant type (r or w) on the same plate because 
of coincidence of two mutations, either in the same bacterium or in the group of 
bacterial bursts examined on one plate. The expected coincidences ("doubles") 
were calculated to be 2.6 r and 4.3 w. It is not easy to correct for these "doubles," 



146 



FREQUENCY OF SPONTANEOUS BACTERIOPHAGE MUTANTS 




32 64 X 



FIG. 1. The distribution of mutant clones, x = number of mutants in a clone. 
Y = number of clones with x or more mutants. X : r mutants. C : w mutants. 
• : r and w mutants. The solid lines represent the theoretical distributions, with 
slope —1. 



147 



S. E. LURIA 

which cannot be recognized by inspection. A first approximation can be made by 
assuming that all "doubles" include a clone of one mutant, and that they all 
occur among the most frequent clone size classes observed (plates with 2, 3, 4 
mutants). The resulting corrected distribution has a slight excess of ones, but 
does not deviate significantly from Equation (2), in spite of the fact that the 
correction is an extreme one, which concentrates all the distortion in the initial, 
most critical portion of the frequency distribution curve. 

3. Failure of the plaque count method to reveal the full number of mutants is 
unhkely to result from technical reasons, but might be due to intrinsic properties 
of phage reproduction. For example, only a fraction of the gene copies produced 
might appear in active phage particles because of loss or inactivation of a certain 
proportion of phage within the bacteria before liberation. Such "sampling 
losses," superimposed on a logarithmic distribution, would make the initial slope 
of log Y versus log x steeper; a samphng loss of 50 per cent would give a quite 
appreciable deviation from the initial slope of —1. This is in conflict with our 
results. Sampling losses could be superimposed on any other distribution, but it 
is hard to visualize how the results could simulate those of an exponential 
distribution. 

A variety of alternate hypotheses was considered in an attempt to find one, 
besides that of exponential reproduction, that could lead to the results found 
experimentally. No sensible hypothesis could be devised. Altogether, the 
hypothesis of mutations occurring at a constant rate in the course of expoiiential 
nonsynchronized gene reduphcation appears adequate to account for our results. 

Mutation Rates 

Our results permit fairly accurate estimations of mutation rates. A total of 
87.6 r mutations (85 mutant clones plus 2.6 coincidences) occurred in 23,000 
bursts, producing approximately 1,850,000 active phage particles. The mutation 
frequency per reduplication is around 5 X 10~^. 102.3 w mutations in 11,000 
bursts, or 880,000 particles, correspond to a mutation frequency of 1.2 X 10~*. 
These mutation rates, of course, are the sums of the mutation rates at all the 
individual loci that can mutate to give either the r or the w phenotype. 

Incidental Observations 

1. The mutants of T2L classified as r or tt^ are generally clearly recognizable as 
such. The r phenotype, however, is not uniform; different r mutations give 
plaque types often distinguishable from one another. The w phenotype, though 
generally sharply distinct from r and from wild-type, is even more variable. Only 
one clone (consisting of one plaque), even after repeated replatings, could not be 
classified with certainty as either w or r; since at any rate it seemed to represent 
a novel phenotype, it was excluded from the analysis. 

2. Several other types of mutants were observed in the course of our experi- 
ments, mainly "minute" or "sharp" plaque types. These were not included in 
the results. 



148 



FREQUENCY OF SPONTANEOUS BACTERIOPHAGE MUTANTS 

.3. Mottled plaques were often seen on our plates. They generally derive from 
mutations occurring during plaque formation. In several cases, mottled plaques 
were present on the same plate with r mutants. Three such mottled plaques were 
replated and the r component strain was isolated and crossed with a strain from a 
pure r plaque on the same plate. In two cases wild-type recombinants appeared, 
indicating that the mottled plaque probably stemmed from an independent r 
mutation that occurred on the plate; in the third case no recombinant was found. 
The latter type of mottled plaque could appear in the yield of a bacterium 
containing a clone of r mutants, owing to some limitations to the complete 
segregation of virus particles or of their genetic components (Hershey and Chase, 
1951). 

Discussion 

The exponential rate of gene reproduction in phage, suggested by our analysis 
of the clonal distribution of spontaneous phage mutants, simply means that the 
initial gene copy brought in by the infective virus does not possess the monopoly 
of replication. Its copies act in turn as sources for new replications and the 
elementary replication process is a reduplication. This conclusion, if correct, 
represents a step further in our analysis of phage reproduction. 

Our results do not contribute any information as to whether the mutations 
occur only at reduplication or between reduplications, in the "interphase." They 
are compatible both with production of phage particles as such and of individual 
genetic components, although the apparent lack of "sampling losses" is more 
easily reconciled with the former. 

There is some interest in comparing the clonal distribution of spontaneous 
mutants with the almost random distribution of recombinant phage from 
mixed-infected bacteria (Hershey and Rotman, 1949). The very first active 
phage particles that appear inside mixed-infected bacteria include recombinants 
(Doermann and Dissosway, 1949), whose distribution is also intermediate 
between a clonal, reduplicational one and a random one. The facts suggest the 
following conclusions: 

1. Recombination occurs late in reproduction of the genetic material of phage. 
This may be due either to the coincidence of recombination with some terminal 
step in phage maturation or to an increased probability of recombination when 
large numbers of phage elements are present in a cell. 

2. Recombinants detectable as mature particles around the middle of the 
intracellular growth period do not reduplicate as such; otherwise, by giving rise 
to clones, they would cause the later population of recombinants to be distributed 
more and more like spontaneous mutants, contrary to experimental finding 
(terminal distribution of recombinants almost random). 

3. Reproduction of the genetic material of phage, therefore, takes place 
mainly by reduplication of elements that are not yet in the form of mature phage 
particles. 



149 



S. E. LURIA 

References 

Adams, M. H., 1950, Methods of study of bacterial viruses. In: Methods in Medical 

Research. 2: 1-73. 
DelbrIjck, M., and Bailey, W. T., Jr., 1946, Induced mutations in bacterial viruses. 

Cold Spring Harb. Symposium Quant. Biol. 11: 33-37. 
DELBRtJCK, M., and Luria, S. E., 1942, Interference between bacterial viruses. I. Arch. 

Biochem. /: 111-141. 
DoERMANN, A. H., 1948, Intracellular growth of bacteriophage. Yearb. Carneg. Instn. 47: 

176-182. 
DoERMANN, A. H., and Dissosway, C. F.-R., 1949, Intracellular growth and genetics of 

bacteriophage. Yearb. Carneg. Instn. 48: 170-176. 
DuLBECCO, R., 1949, The number of particles of bacteriophage T2 that can participate in 

intracellular growth. Genetics 34: 126—132. 
1952, A critical test of the recombination theory of multiplicity reactivation. J. Bact. 

(in press). 
Hershey, a. D., 1946a, Mutations of bacteriophage with respect to type of plaque. 

Genetics 31 : 620-640. 
1946b, Spontaneous mutations in bacterial viruses. Cold Spring Harb. Symposium 

Quant. Biol. 11: 67-76. 
Hershey, A. D., and Chase, M., 1951, Genetic recombination and heterozygosis in 

bacteriophage. Cold Spring Harb. Symposium Quant. Biol. 16. 
Hershey, A. D., and Rotman, R., 1948, Linkage among genes controlling inhibition of 

lysis in a bacterial virus. Proc. Nat. Acad. Sci. Wash. 34: 89-96. 
1949, Genetic recombination between host-range and plaque-type mutants of bacterio- 
phage in single bacterial cells. Genetics 34: 44-71. 
Lea, D. E., and Coulson, C. A., 1949, The distribution of the numbers of mutants in 

bacterial populations. J. Genet. 4-9: 264-285. 
Luria, S. E,, 1945a, Mutations of bacterial viruses affecting their host range. Genetics 30: 

84-99. 
1945b, Genetics of bacterium-bacterial virus relationships. Ann. Missouri Bot. Gard. 

32: 235-242. 
1947, Reactivation of irradiated bacteriophage by transfer of self-reproducing units. 

Proc. Nat. Acad. Sci. Wash. 33: 253-264. 
Luria, S. E., and DEUBRticK, M., 1943, Mutations of bacteria from virus sensitivity to 

virus resistance. Genetics 28: 491-511. 
Luria, S. E., and Dulbecco, R., 1949, Genetic recombinations leading to production of 

active bacteriophage from ultraviolet inactivated bacteriophage particles. Genetics 

34: 93-125. 
Luria, S. E., and Human, M. L., 1950, Chromatin staining of bacteria during bacterio- 
phage infection. J. Bact. 59: 551—560. 
Luria, S. E., Williams, R. C, and Backus, R. C, 1951, Electron micrographic counts of 

bacteriophage particles. J. Bact. 61 : 179—188. 

DISCUSSION 

Altenburg: I suggest that the conclusion to be drawn from Dr. Luria's 
experiments is that the reproduction cycle of viruses conforms with that of 
organisms in general, and that therefore viruses are to be considered organisms. 



150 



GENETIC RECOMBINATION BETWEEN HOST-RANGE AND 

PLAQUE-TYPE MUTANTS OF BACTERIOPHAGE IN 

SINGLE BACTERIAL CELLS^-^ 

A. D. HERSHEY and RAQUEL ROTMAN^ 

Department of Bacteriology and Immunology, Washington University Medical School, 
St. Louis, Missouri 

Received June 28, 1948 

WE HAVE previously shown that any two of several independently 
arising plaque-type {r) mutants of the bacterial virus T2H interact with 
each other, in bacterial cells infected with both, to give rise to wild type and 
double mutant genetic recombinants (Hershey and Rotman 1948). In this 
paper we describe comparable interactions between host-range and r mutants 
of the same virus. The experiments furnish new information because it has 
proved possible to count the numbers of all four types of virus found in yields 
from the mixedly infected bacteria. 

MATERIALS AND METHODS 

The types of viral mutant to which we shall refer in this paper may be 
summarized in terms of the mutational pattern illustrated in fig. \. In this 
diagram, /; refers to a host-range mutant, r to any one of the rapidly lysing 
mutants (Hershey and Rotman 1948), and m ("minute") to a mutant not 
previously described which is characterized by a very small haloless plaque. 
The h mutant is one which forms plaques identical in appearance and number 
on typically sensitive strains of Escherichia coli, and on an indicator strain 
(No. 2 B/2H, 2K) resistant to h+ forms of the virus (Hershey 1946a). All the 
steps indicated in the diagram by arrows can be observed either as spontaneous 
mutations, or by making the appropriate crosses. Only one example of the 
mutant w, obtained by crossing wild type with an rm arising in a stock of the 
mutant rl3, has been studied. The plaques of m and rm are different, but are 
not easily distinguishable, as shown in the photograph (fig. 2). 

In principle, the experimental technique we have to describe is very similar 
to that of genetic crossing, and will be referred to in this paper in genetic 
terms. One starts with a pair of mutants, each corresponding to a mutant 
haploid germ cell differing from wild type by a different unit change. Bacterial 
cells are infected with both members of the pair, and during viral growth the 
pair interact to produce viral progeny corresponding to germ cells of a new 

' Aided by a grant from the U. S. Public Health Service. 

* The manuscript was prepared while the senior author held a temporary appointment in the 
Department of Biology of the California Institutk ok Technology. It is a pleasure for him to 
acknowledge material and intellectual aid received from members of the staff of that department. 

' Present address: University of Minnesota, Minneapolis, Minn. 



Reprinted by permission of the authors and Genetics, Inc. from 
Genetics, 34, 44-71, January, 1949. 

151 



RECOMBINATION IN BACTERIOPHAGE 



45 



generation, but now including some individuals differing from wild type by 
both unit changes, and other individuals differing from wild type not at all. 




Kr mr 

Figure 1. — Mutational pattern of the bacterial virus T2H. 

The analogy to other genetic recombinations is obvious, and it is natural to 
look for a common mechanism. 

The procedure of making a cross consists essentially in infecting a measured 
number of growing bacteria with larger measured numbers of two kinds of 
virus, diluting the culture before lysis begins to prevent readsorption of viral 
progeny to bacteria not yet lysed, and plating samples of the total yield of 









• 



I 

i 



Figure 2. — Progeny of the cross mY.rl. 



152 



46 A. D. HERSHEY AND RAQUEL ROTMAN 

virus for a differential count of its component types. This procedure was first 
used, for another purpose, by Delbruck and Luria (1942). 

The advantage of the /;Xr cross is that all the genetic types of virus to 
which it gives rise can be recognized in a single plating on a mixture of bac- 
terial strains (fig. 3). This makes possible the analysis of viral yields from single 
bacterial cells. For this purpose the procedure already described is modified 




^ 



4 ' ,/ 



I 



• 




ii 




• • '' 


• •' 


• 


r 
ht - 


; 


m 



Figure 3. — Progeny of the cross hy^rl plated on mixed indicator. The acentric clearings in 
the k^r plaques result from secondary h mutations. 

by increasing the factor of dilution to obtain only one infected bacterium to 
about three ml of nutrient broth. The culture is then divided up into samples 
of one ml, most of which will contain either no bacteria at all, or a single one. 
A large number of these samples are plated out after the bacteria lyse, and the 
elementary yields, averaging about 500 particles of virus in our experiments, 
are analyzed in toto. Both the mixed indicator method for differential counting 
of viral mixtures, and the single burst technique as here employed, were first 
used by Delbruck (1945a, b). 

The remaining portion of this discussion of materials and methods is of 
technical interest only. 

Viral stocks are prepared by seeding nutrient broth cultures of E. coli strain 
S with material taken directly from a single plaque. Stocks of wild type and 
h mutant usually have titers exceeding 10^ V™!; '' rnutant titers are always a 



153 



RECOMBINATION IN BACTERIOPHAGE 47 

little less. Many of the stocks, particularly of the r mutants, contain unknown 
substances inhibiting adsorption of the virus if the dilution into bacterial 
suspension is less than 1 : 100 or so. There is no difference in the adsorption of 
different stocks at high dilutions of the virus. If a mixed infection is attempted 
with an r stock which contains inhibitor, and an h or wild type stock which 
does not, the adsorption of both viruses in the mixture is prevented equally. 
No undesirable disturbance of the relative multiplicity of infection in mixtures 
is therefore encountered. To obtain satisfactory levels of adsorption, all r 
mutant stocks are sedimented in a refrigerated International centrifuge with 
Multispeed head, and resuspended in a solution containing 1 percent Bacto- 
peptone and 0.5 percent NaCl. Very little loss of virus occurs during sedimen- 
tation, and probably no permanent aggregation, since sedimented mixtures of 
r and wild type virus do not yield mixed plaques. The resuspended virus is 
stable for at least several weeks. 

To make a cross, a two hour culture of E. coli strain H in nutrient broth, 
containing 2X10^ bacteria per ml, is infected at 37°C in an aerated culture 
tube (the "adsorption tube") with yV volume of a mixture of diluted viral 
stocks containing 2X10^ plaque forming particles of each kind per ml. After 
five minutes, during which equal numbers (about 50 percent) of each virus 
are adsorbed, a 10* dilution is made into broth (the "growth tube") for further 
incubation, and a second diluted sample is spun for the assay of unadsorbed 
virus. Sixty minutes after infection, an assay from the growth tube gives the 
average yield of virus from about 40,000 mixedly infected bacteria. 

For the single burst experiments, an additional dilution from the adsorption 
tube into antiserum to neutralize the unadsorbed virus (Delbruck 1945c), 
is made at the end of the adsorption period. Five minutes later, a further 
dilution from the antiserum tube is made into broth to contain about one 
infected bacterium per three ml. Before the 20th minute after infection, sam- 
ples measuring one ml are distributed into a series of small tubes. The virus 
yields in these tubes are assayed by plating 0.3 ml 60 minutes or more after 
infection. The remainder of each sample, excepting those containing no virus 
or unmixed yields, is plated on two additional plates the next day. About 10 
percent of each sample is mechanically lost. An important feature of these 
experiments is the guard against contamination of materials provided by the 
fact that about f of the samples contain no virus, whereas the remainder 
contain more than 100 particles. 

Viral yields from the growth tube are plated on sensitive bacteria (strain 
S), on the indicator strain (No. 2 B/2H, 2K), and on a mixture containing 
one volume S and two volumes indicator (day-old broth cultures). On the 
mixed indicator all four types of virus can be recognized (fig. 3), and their 
sum equals the count on S. Mixed indicators plates always show a few doubt- 
ful plaques which can be identified only by sampling and retesting, but their 
number is too small to be of importance, and mixtures of pure stocks can be 
counted with satisfactory accuracy. The counts on the single indicator, giving 
only the h mutants, are also satisfactory with mixtures of pure stocks. These 
counts tend to be low, however, for mixed yields of h and A+ virus. The cause 



154 



48 A. D. HERSHEY AND RAQUEL ROTMAN 

of this "mixed indicator effect," similar to that observed by Delbruck and 
Bailey (1946), remains obscure. It does not appear to be the result of segre- 
gation of multiple h factors, because no intermediate genetic types can be 
found in crosses between h and wild type. In h)>ir crosses, it affects equally 
counts of parental and recombinant virus. 

The validity of the mixed indicator count itself rests on three lines of evi- 
dence. First, plaques sampled and retested always conform to the genetic type 
deduced from inspection. Second, the ratio of // to k^ virus in the yield following 
mixed infection, measured by the mixed indicator count, is the same as the 
corresponding ratio of infecting viruses, with minor exceptions to be mentioned 
below. Third, the yield of the two recombinants in either /fX?* or hr X wild type 
crosses is very nearly equal, and the slight bias (actually of doubtful signifi- 
cance) correlates not with the h, but with the r pair of alleles. 

For these reasons, and because of its statistical efficiency, only the mixed 
indicator plating is used for counting single bursts. The reproducibility of the 
counts so obtained may be judged from the examples shown in table 1. 

Table 1 

Mixed Indicator Counh of Viral Types in the First Eight 
Bursts Examined from the Cross hXr7. 

The counts shown are for 3 aliquots of 0.3 ml from each tube. The volumes are not measured 
very accurately, owing to the effort made to plate the entire sample. In computing results, it was 
assumed that totals of counts for each tube represented 90 percent of the actual virus content, 
10 percent of the fluid being lost mechanically. 

TUBE /z+r"*" hr'^ h^r hr tube h'^r'^ hr^ h'^r hr 

35 



10 18 17 95 13 36 



9 


57 


46 


19 


9 


77 


56 


20 


11 


50 


47 


18 


18 


17 


95 


13 


10 


14 


73 


8 


10 


22 


118 


15 


9 


84 


51 


8 


10 


124 


53 


5 


8 


78 


34 


7 


2 


20 


58 


17 


6 


26 


78 


32 


6 


ii 


55 


16 



26 9 84 51 8 38 



34 2 20 58 17 40 



21 


80 


71 


23 


24 


74 


65 


18 


14 


68 


39 


15 


9 


136 


30 


9 


7 


100 


22 


7 


14 


120 


18 


9 


9 


139 


30 


9 


9 


130 


19 


13 


4 


102 


31 


4 


1 


24 


54 


1 


3 


38 


83 


10 


3 


24 


66 


7 



The nutrient broth referred to above is composed of Bacto-peptone 10 g, 
Bacto-beef extract 3 g, NaCl 5 g, glucose 1 g, per liter distilled water. The 
pH (unadjusted) is about 6.8. An occasional batch of broth prepared according 
to this formula proves unfavorable to the adsorption of the virus. 

Nutrient agar plates are poured with a minimum of 35 ml per 9 cm Petri 
dish of Bacto-agar 10 g, Bacto-Tryptose 10 g, NaCl 8 g, sodium citrate crys- 
tals 2 g, glucose 1 g, per liter distilled water. The pH of the agar is adjusted 
to 6.8 to 7.0. Contrary to an early experience, we have recently found that 
Bacto-Tryptone can be substituted for Tryptose. Poured plates are stored in 



155 



RECOMBINATION IN BACTERIOPHAGE 



49 



the refrigerator. Agar for layer plating has the same nutrient composition, 
but contains only 0.5 to 0.6 percent agar, the optimal concentration depending 
on the age of the plates and on other variables. 

All platings of virus are made by the agar layer method, by adding an aliquot 
of virus measuring 0.1 to 0.5 ml and about 0.1 ml of a day old unaerated broth 
culture of bacteria, to 2 ml of melted soft agar at 45°C, and pouring the mix- 
ture over the surface of an agar plate at room temperature. The plates are in- 
cubated 18 to 24 hours at 37°C without inverting, during which time fluid 
collects on the agar surface without undesirable effects if conditions are 
optimal. 

Bacterial counts are made by spreading 0.1 ml aliquots on the surface of 
dried agar plates. 

THE LINKAGE SYSTEM 

The mutants m and h {minute plaque and host-range modification, respec- 
tively) have been crossed with wild type, with the mutants called rl, r7, and 
rl3, and with each other. The crosses with wild type yield only the parental 
types of virus, confirming that the mutants are unit modifications. The crosses 
with r mutants, and the intercross, extend the linkage system previously de- 
scribed (Hershey and Rotman 1948) as shown in fig. 4. The h locus is closely 
linked to the locus rlJ; the m locus belongs in a third linkage group C. 

It should be understood that the diagram of fig. 4 is only a convenient 
representation of linkage relations whose structural basis remains to be eluci- 
dated (Hershey and Rotman 1948). The question of linear structure will be 
returned to in the discussion of this paper. At this point we insert an experi- 
ment which supports the general interpretation in terms of linkage. 



A 



Pi ri4 
— H — 



B 



1% 

-f+— 



3. 



\ 



yi57o 



r 2, 4, 8, 9, 3, 6, 7, 5 
I I I I I I I I 



^ 



7% 



J 



y 



15% 



Vl5% 



m 



y 



Figure 4. — Linkage relations among mutants of T2H. The percentages indicate 
yields of wild type in two factor crosses. 

The double mutants h rl and h rl3 were isolated by making the respective 
crosses hy,r7 and hXrl3. The three crosses r7Xrl3, hr7Xrl3, and hrl3Xr7 
were then compared with respect to the yield of r+ virus, classified without 



156 



50 A. D. HERSHEY AND RAQUEL ROTMAN 

respect to the host-range character. In each case the yield was the same with- 
in experimental error, seven percent of the total virus. This experiment shows 
that the small yield of recombinants (one percent of wild type) in the cross 
hXrlJ cannot be attributed to suppression of a hypothetical conjugation, 
which would affect the interaction of h rl with rl3 as well, but indicates some 
kind of linkage between the genetic factors concerned. 

AVERAGE YIELDS OF VIRUS IN CROSSES BETWEEN HOST-RANGE AND r MUTANTS 

Six crosses have been studied by the single burst technique; namely, hX.rl, 
hy,r7, hXrJJ and the corresponding reverse crosses, h rX wild type. Each 
cross was made three to five times for the collection of the single burst data, 
and each time the viral yield was examined also from a culture tube containing 
about 40,000 mixedly infected bacteria. The average yields from the culture 
tubes are summarized in table 2. They show that the h locus is very closely 
linked to rl3 (less than one percent of wild type), and that the linkage rela- 
tions to rl and r7 are approximately what would have been predicted from this 
fact, respectively 12 and 6 percent of wild type. According to arguments 
previously given (Hershey and Rotman 1948), the factors /;, r7, and rl3 are 

Table 2 

A verage Percent Distribution of Viral Types in Yields 
from about 40,000 Mixedly Infected Bacteria 

The results shown are from the same experiments for which single burst data are also re- 
ported, except that one growth tube was lost among the crosses hr7 by wild type. The total multi- 
plicity of infection is about five of each type per bacterium; the adsorption period is five minutes; 
the total incubation period one hour. The distribution of viral types is computed from the re- 
sults of mixed indicator platings. The column headed eop(h) gives the efficiency of plating of h 
virus on single indicator as compared with mixed indicator, and illustrates the mixed indicator 
effect mentioned in Methods. The column headed p(h) gives the percent of virus containing the h 
allele, and shows the effect of selection during growth. 



NO. OF 
EXPTS. 



/?+r^ 



//r^ 



h^r 



hr 



eop(h) p(h) 



hXrl 


5 


input 


53 


47 







1.0 


53 






yield 12 


42 


34 


12 


630 


0.8 


54 


//riX-|- + 


3 


input 57 








43 




1.0 


43 






yield 44 


14 


13 


29 


680 


0.7 


43 


//X/-7 


4 


input 


49 


51 







1.1 


49 






yield 5.9 


56 


32 


6.4 


650 


0.8 


62 


hr7X^ + 


2 


input 49 








51 




1.0 


51 






yield 42 


7.8 


7.1 


43 


690 


0.8 


51 


liXrlJ 


3 


input 


49 


51 







1.1 


49 






yield 0.74 


59 


39 


0.94 


510 


0.8 


60 


hrl3X + + 


4 


input 52 








48 




0.9 


48 






yield 50 


0.83 


0.76 


48 


590 


0.7 


49 



157 



RECOMBINATION IN BACTERIOPHAGE 



51 



linked to each other, but rl is probably situated on an independently ex- 
changing structure. 

The results of table 2 show further that the two recombinants appear in 
equal numbers in any one cross, and that pairs of reverse crosses yield equal 
numbers of recombinants. It is these relations, which increase the resemblance 
to simple types of Mendelian segregation, that we wish to examine by the 
single burst technique. In the remainder of the experimental part of this 
paper we describe the results of this examination, but limit our comments 
chiefly to the technical problems encountered. The general implications of the 
data will be considered in the discussion. 

VARIATIONS IN YIELDS OF PARENTAL TYPES OF 
VIRUS AMONG INDIVIDUAL BACTERIA 

Yields of virus from single bacteria show large fluctuations in size (Del- 
BRiJCK 1945b) and, in our experiments, variations in relative yields of the 
two infecting viruses. The variations in total yield are shown in fig. 5 which 
includes the complete data for six experiments in which the proportion of 
multiple bursts is small (100 bursts out of 484 tubes, or 11 probable multi- 
ples). The burst sizes range from 150 to several thousand, with a mean of 
520, or 470 corrected for probable multiples. The distribution is the same for 
the mixed bursts, and for the bursts containing only one viral type. Owing 
to these variations, it is convenient to describe the individual bursts in terms 
of the fractional yield of the several viral types. 




400 500 600 
Burst size 



>1000 



Figure 5. — Distribution of total viral yields among 100 single bursts from mixedly infected 
bacteria. The solidly shaded areas refer to unmixed yields. 

Each cross yields four types of virus, wild type, the single mutants h and 
r, and the double mutant h r, of which two are parental and two are recombi- 
nant types. In order to examine the variations in relative yields of parental types 
independently of variations in yields of recombinants, it is convenient to ex- 
press the former in terms of the proportion of virus containing a specified 
allele, viz: 



158 



52 



. D. HERSHEY AND RAQUEL ROTMAN 




n(/zr+) + n(/?r) 

p(/?.) = 

n (total) 


(1) 




(2) 



n (total) 

where n indicates the number of the specified viral types, p{h) is the propor- 
tion of the yield containing the h allele, and 1 — p(/f), p(r), and 1 — p(r) are re- 
spectively the proportions containing the //+, r, and r+ alleles. The fraction 
p(^) is equal to 1 — p(r) if the numbers of the two recombinant types are the 
same, or approximately equal if these numbers are small. A fair idea of the 
complete distribution of alleles is therefore given by the distribution of p(/f) 
alone. Two examples of this distribution, showing the proportions of the h 
allele in diflferent single bursts for the crosses hXrl and hXr7, are given in 
fig. 6. 



01 lo- 



co 

O 

6 

d 




18 27 36 45 54 63 72 81 90 93 100 

per cen-b K virus 



10 

M-. 
O 

L, 

s 




27 3S 45 54 63 72 81 90 

per cent K virus 

Figure 6. — Distribution of proportions of It virus in yields from single bacteria. 



100 



159 



RECOMBINATION IN BACTERIOPHAGE 53 

The variations shown in fig. 6 are evidently due in part to variations in 
relative numbers of the two kinds of virus adsorbed, and in part to variations 
in viral growth. For purposes of comparison, the variation in relative multi- 
plicity has been computed on the assumption of a random distribution of two 
types of virus over a population of bacteria receiving on the average five 
particles of each kind. The distribution for this case is approximately that 
shown in table 3. 

Table 3 
Ideal Dislribiition of Multiplicities in Mixed Infection 

The proportions of l^acteria falling into the specified groups classified with respect to relative 
multiplicity of infection with two kinds of virus have been calculated for random adsorption with 
average multiplicity of five each of the two kinds. 



NUMBERS OF FIRST KIND 
AS PERCENT OF TOTAL VIRUS PERCENT OF BACTERIA 

ADSORBED 



or 100 
1-9 or 91-99 
10-18 or 82-90 
19-27 or 73-81 
28-36 or 64-72 
37-45, 46-54, or 55-63 
total 



0.7 


(exact) 




0.2 


(approximate) 


2.0 


u 




5.1 


u 




10.8 


u 




18.2 


" 




92.2 







The distributions of yields actually found (fig. 6) differ from the theoretical 
distribution of multiplicities in showing a considerably broader spread, and a 
significant excess of yields containing only one kind of virus. These effects 
could be due in part to an inhomogeneity of the bacteria with respect to ad- 
sorbing power for virus; otherwise they suggest that a bacterium infected with 
two viruses is somewhat less likely to liberate a given one, than a bacterium 
infected with that one alone. Dulbecco (1949) has shown that the latter is, 
in fact, the case. 

Another possible contribution to the variations described is connected with 
the relatively long period (five minutes) allowed for adsorption of virus, which 
permits some bacteria to be infected with one or more particles of one type of 
virus considerably in advance of infection by the second. Owing to the slow 
adsorption of the virus T2H and its mutants, the adsorption time cannot be 
much reduced without reducing the total multiplicity of infection, or introduc- 
ing excessive amounts of virus. This contribution to the variation in composi- 
tion of viral yields has not, therefore, been assessed. 

COMPETITION BETW'EEN VIRAL MUTANTS 

The competition between viral mutants expresses itself in two ways; first, 
by the complete suppression of one virus or the other in mixedly infected bac- 
teria, and second, by excessive growth of one of the two types in bacteria liber- 
ating both. These effects are slight among mutants of T2H so far examined, 



160 



54 A. D. HERSHEY AND RAQUEL ROTMAN 

and probably do not influence appreciably the yields of recombinants in genetic 
crosses, as the following discussion will show. 

The two distributions shown in fig. 6 illustrate the competitive relations 
encountered. In the cross hy,r7, most of the unmixed yields contain the h 
rather than the r parental type. Corresponding to this, there is a tendency for 
the mixed yields to contain an excess of h virus. The combined effect is to cause 
a definite increase in yield of h virus at the expense of r, as compared with the 
input proportions. The cross hXrlJ also shows these characteristics, the ef- 
fects being evident in table 2. 

In the cross hy,rl, on the other hand, the unmixed yields of each kind are 
approximately equal in number, and the mean proportion of h virus in the 
mixed bursts, in the total yield, and in the input mixture of viruses is the 
same. The crosses between wild type and hrl, hr7, and hrl3 are like AXr7 in 
this respect, as shown in table 2. 

It might be supposed that the suppression of one virus by a second is favored 
by an excess of the second. This is true only in a special sense, as Dulbecco 
(1949) has shown, and we have confirmed. An excess of one virus tends to sup- 
press a minority type completely in some bacteria, but there is a compensating 
excess of this type among the mixed bursts, so that the average proportion of 
the minority virus in the yield averaged over many bacteria is the same as in 
the input mixture. This identity has been established with considerable pre- 
cision for proportions of rl between 7 and 50 percent in mixed infection with 
wild type. The nature of this relationship, which is at first sight perplexing 
in the case of unequal multiplicity, has been explained by Dulbecco (1949) 
in terms of a limitation to the number of viral particles which can participate 
in growth in a single bacterium. If all those viral particles in excess of a certain 
number attached to the same bacterium fail to grow, and if the excluded ones 
are chosen at random, the result will be precisely the one described, provided 
there is no selection during the growth of the successful particles. 

It is apparent that with certain viral pairs, the excluding mechanism does 
not operate at random, or there is continuing selection during growth. Thus 
h mutant slightly suppresses r7 or rl3, but not rl or wild type. Wild type 
suppresses rl3, but not rl or r? (Hershey and Rotman 1948). There is no 
selection with respect to either h or r factors when wild type is crossed with 
h rl, h rJ, or h rl3 (table 2). 

The competitive relations discussed above are of immediate interest only 
in the negative sense that they probably do not influence the yields of recom- 
binant virus in crosses. The evidence for the latter conclusion, drawn from data 
presented elsewhere in this paper, may be summarized as follows: (1) the link- 
age relations deduced from average yields of virus are the same as those de- 
duced from single bursts selected for equality of yields of the two infecting 
viruses; (2) in the reverse crosses hy.r7 and h r7 Y, wild type, one gets with- 
in experimental error equal numbers of all four recombinants in spite of the 
fact that in one case the infecting pair, and in the other the recombinant pair, 
have unequal excluding power; (3) in all crosses, the distribution of yields of 
recombinants among single bursts does not show one peak at zero and another 



161 



RECOMBINATION IN BACTERIOPHAGE 55 

above the mean, as is the case with the yields of a minority infecting type, 
but shows a single mode slightly less than the mean. 

The last two lines of evidence cited seem to show that the principle of limited 
participation (Dulbecco 1949) referred to above, operates only during the 
initial stages of infection, or at any rate does not influence the yields of genetic 
recombinants arising within the mixedly infected bacteria. They suggest 
further that the h mutant is superior to r7 or rl3 in excluding power only, not 
as a competititor during actual multiplication. 

YIELDS OF GENETIC RECOMBINANTS FROM SINGLE 
MIXEDLY INFECTED BACTERIA 

In order to study the variations in yields of recombinants intrinsic to the 
recombination process, one would like to exclude as many as possible of the 
accessory sources of variation. The most important of these are variations in 
burst size, and variations in the relative numbers of the two infecting viruses 
adsorbed to individual bacteria. It will be seen presently that effects of varia- 
tions in burst size can be avoided by the simple expedient of computing propor- 
tionate yields of recombinants, these being independent of burst size. The effect 
of variations in relative multiplicity could be minimized either by going to 
very small or very large total multiplicities. Low multiplicities are uneconomi- 
cal, because at multiplicities sufficiently small so that most of the mixedly 
infected bacteria receive only one viral particle of each type, very few of the 
test cultures will yield a mixed burst. High multiplicities also introduce dif- 
ficulties (Dulbecco 1949). We have chosen to use total multiplicities between 
10 and 20, within which range the yield of recombinants is constant. 

As previously described, the elementary viral yields vary considerably in 
the relative numbers of the two parental types of virus and, as expected, these 
variations influence in turn the yields of recombinants. A correction for this 
source of variation was devised as follows. Assuming that the genetic inter- 
action occurs between unlike viral pairs, and that the composition of the viral 
yield provides a direct measure of the composition of the intracellular viral 
population during growth, one computes an interaction coefficient 

k = p(/0[l -pW] (3) 

in which p(h) is given by (1), and k expresses the influence of the composition 
of the population on the number of unlike viral pairs present in the cell, neg- 
lecting effects of genetic recombination. 

The coefficient k has a maximum of 0.25 when half the viral yield contains 
the h allele. Dividing the proportions of recombinants by 4k serves therefore 
as a correction for inequality of yields of the parental viruses. This correction 
is ambiguous only for bursts in which the yields of the two recombinants are 
large and unequal, and bursts from which either recombinant is absent. 

A summary of the single burst data is given in table 4, which includes the 
mixed bursts only. The bursts have been separated into the classes k^0.21 
and k^0.20, to show the effect of the correction described above. It will be 
seen that the uncorrected mean proportion of recombinants is larger for the 



162 



56 A. D. HERSHEY AND RAQUEL ROTMAN 

bursts with the larger k, and that the effect of dividing each proportion by 4k 
is to make the results homogenous. These facts justify the use of the correction. 
Its theoretical significance is clarified in the discussion. 

Table 4 

Single burst data for ItXr crosses 

k=a measure of disproportion between yields of parental types (Eq. 3). 

X =average yield of the r^ recombinant as percent of total virus. 

y = average yield of the r recombinant as percent of total virus. 

n=average burst si/e. 

■■(x, y) =coefficient of correlation between proportionate yields of the two recombinants. 

r{n, x+y) =coefficient of correlation between burst size and sum of proportions of the two recombinants. 

The variations of x, y and n/IOO shown are standard deviations within the sample. The standard errors of the means are 

obtained by dividing these by the square root of the number of bursts. 
The correction referred to is described in the text. 



CROSS 


NO. OF 
BURSTS 


k 


4k 




X 


y 


n/100 


r(.x, y) 


r(n, x+y) 


hxn 


25 


SO. 20 


0.65 


uncorr. 
corr. 


8.9±3.8 
13.9+5.3 


12.116.2 
19.019.2 


5.512.4 


0.09 


-0.25 




52 


§0.21 


0.94 


uncorr. 
corr. 


15.5±6.9 
16.6+6.3 


17.117.1 
18.117.2 


5.813.2 


-0.01 , 


0.15 


lirlX + + 


19 


SO. 20 


0.67 


uncorr. 
corr. 


13. 6 ±5. 9 
20.1+7.4 


13.214.4 
20.818.5 


5.111.9 


-0.27 


-0.26 




36 


go. 21 


0.94 


uncorr. 
corr. 


16.515.8 
17.5 + 6.2 


17.115.6 
18.216.0 


5.4 + 3.0 


0.16 


-0.11 


I'Xr? 


13 


SO. 20 


0.64 


uncorr. 
corr. 


5.7±3.5 
8.8+4.8 


6.6+4.6 
9.915.9 


6.21 1.9 


0.65 


-0.43 




35 


§0.21 


0.93 


uncorr. 
corr. 


6.8±4.2 

7.2±4.3 


9.214.5 
9.8+4.8 


5.913.2 


0.53 


0.16 


h r7X + + 


17 


SO. 20 


0.63 


uncorr. 
corr. 


6.4±3.1 
10.0+4.3 


4.812.9 
8.015.0 


6.417.8 


0.12 


-0.45 




26 


60.21 


0.95 


uncorr. 
corr. 


6.7±3.2 
7.113.6 


8.915.4 
9.315.6 


5.813.7 


0.47 


0.02 


hy.rl3 


29 


SO. 20 


0.61 


uncorr. 


0.510.5 


0.810.8 


5.2 + 2.1 


0.07 


0.35 




29 


60.21 


0.94 


uncorr. 


0.910.9 


1.110.9 


5.112.3 


-0.05 


-0.30 


lirl3X + + 


21 


SO. 20 


0.69 


uncorr. 


0.610.7 


0.610.9 


4.61 1.4 


0.15 


0.22 




46 


60.21 


0.94 


uncorr. 


0.8+0.7 


0.710.6 


4.912.3 


0.21 


0.04 



The data of table 4 for mixed single bursts confirm fairly well the average 
data of table 2, except that the yields of recombinants are somewhat greater 
owing to the exclusion of the unmixed bursts, and that the yields corrected 
for unequal growth of parental viruses are higher still. 

The chief point of interest is the question of the correlation between yields 
of the two recombinants in single bursts. This has been measured in terms of 
the correlation coefficient r (Rider 1939). This measure varies between — 1 and 
+ 1, a value near indicating independence of variates, and values near unity 
indicating negative or positive correlation, respectively. The data of table 4 
show clearly that there is no significant correlation between the proportions 
of the two recombinants in single bursts except for the crosses liXr7 and // r7X 
wild type. Even for these crosses the correlation is weak and not entirely 
convincing, especially since the data are not completely unselected (see below). 



163 



RECOMBINATION IN BACTERIOPHAGE 57 

The correlation between the uncorrected proportions of the two recombi- 
nants is shown in the form of scatter diagrams in fig. 7. These data might be 
expected to show some degree of spurious correlation owing to the fact that 
bursts with disproportionate yields of the two parental types tend to contain 
diminished numbers of both recombinants. This tendency can be seen in the 



44 1 % h*r^ 

40 1 

28' 



:f 



h X rl 



24 
20- 
16 
12- 

8 
4 
0^ 









•8Sf • . rp 



• o o 



o ••• 

X 

%hr 



4 8 12 16 20 24 26 32 36 

hx rl3 
4 

3 






O o' o 



o 







i»-«X0—O0 r 

1 2 



28i % hr'' 

24- 



hrlxh*r* 



8 •© 



oo 



• • 



• o 



4 



%hr" 



12 16 20 24 2S 32 



hrl3xK*r* 



• •• 

>o?Pa* 



% h^r 



3 



22i % h.-'r' 

16' 

)4-| 

12 

10 



hx r7 






z • 



« 



%hr 



2 4 6 



10 12 14 16 



:iV 



34 



hr7x h^r' 



14, %hr* 

12- 

10- 






%KV 



2 4 6 



10 12 14 16 18 22 Z4 



Figure 7.— Correlation between proportionate yields of the two recombinants in single bursts. 
The open circles indicate bursts with disproportionate yields of the two parental viruses. Crosses 
indicate yields omitted from the data of table 4. 



164 



58 A. D. HERSHEY AND RAQUEL ROTMAN 

diagrams in which the disproportionate yields (k^0.20) are indicated by open 
circles. Actually this effect is of minor importance, since the variations due to 
unknown causes are so much greater than those due to variations in k. Con- 
sequently, the uncorrected data lead to the same conclusions as the corrected 
data; namely, that the proportionate yields of the recombinants are uncor- 
related in the crosses hXrl, hXrlJ, and the corresponding reverse crosses, but 
that there is a weak positive correlation for the crosses hXr7 and h r7X wild 
type. 

As mentioned earlier, the correlation data in table 4 are not unselected. The 
diagrams of fig. 7, however, show all the mixed bursts for the respective ex- 
periments, those omitted from the table being indicated by crosses in the dia- 
gram. It is plausible that some of the discrepant bursts came from bacteria 
infected with a spontaneous mutant present in one of the parental stocks of 
virus. In fact, two bursts from the cross hXr7 were found to contain a large 
proportion of mutants, m in one case and weak inhibitor (Hershey 1946b) in 
the other, which almost certainly arose in this way. The h stock used in these 
crosses contained about 0.1 percent of A r virus, so that in crosses with r at 
least one bacterium in 200 was infected with three types of virus, and would 
be likely to yield an excess of one of the recombinant phenotypes. Unfortu- 
nately the recombinant progeny in the exceptional bursts were not checked by 
crossing with the parental stocks, which should be done in any further experi- 
ments of this type. For the present these bursts throw some doubt on the sig- 
nificance of the positive correlation between proportions of recombinants in 
the crosses hXr7 and hr7X wild type. 

The results of the cross hXrl3 are of special interest because of the small 
yield of recombinants. The distribution of the bursts with respect to absence 
of recombinants is shown in table 5. Nine of the 125 mixed bursts fail to show 
either recombinant, and 31 more lack one recombinant or the other. One can 
test the hypothesis that the two sister recombinants arise independently as 
follows. About 20 percent of the bursts fail to show a specified one of the two 
possible recombinants. If the absence of the one were independent of the ab- 
sence of the other, about (0.2)2 qj. fQ^. percent of the bursts should show 
neither recombinant. The number found, 9/125, is larger than this, but not 
significantly so. Moreover, bursts lacking one recombinant do not show less 
than the average proportion of the other (table 5). The data evidently fail to 
exclude the hypothesis of independent origin of the two recombinants, but do 
not, of course, rule out the hypothesis of reciprocal exchange. 

Another question that arises in connection with the data is concerned with 
the number of genetic exchanges per bacterium. For explicitness, we consider 
separately the hypotheses of reciprocal and non-reciprocal exchange. If ex- 
changes are reciprocal, the bursts lacking a single recombinant are the result 
of failure to recognize the few plaques of either type, of losses in the ten per- 
cent of each culture not examined, and of unspecified biological accidents. As 
previously computed, four percent or 5 of the 125 bursts fail to show either 
recombinant for one or another of these reasons, leaving only 4 without re- 
combinants possibly owing to failure of exchange. This number is too small 



165 



RECOMBINATION IN BACTERIOPHAGE 59 

if some of the single absences result from erroneous recognition of one recom- 
binant, but is otherwise subject only to its sampling errors. Taken as a measure 
of failure of reciprocal exchange, the fraction 4/125 implies an average of 
3.4 exchanges per bacterium. If, on the other hand, exchanges are not recipro- 
cal, the fraction (20 percent) of the bursts lacking any given recombinant cor- 
responds to 1.7 genetic transfers per bacterium. 

These estimates may be too low if conditions vary from bacterium to bacte- 
rium in such a way that genetic exchange is suppressed in some bacteria. 

Table 5 

Distribution of Bursts with Rtspect to Absence of Recombinants 
See legend table 4. The proportions of recombinants have not been corrected for dispropor- 
tionate yields of parental types of virus. 



CROSS 


CLASS 


NO. OF 

BURSTS 


X 


y 


n 


k 


hXrlJ 


both absent 


2 


0.0 


0.0 


390 


0.20 




h'^r'^ absent 


10 


0.0 


0.81 


460 


0.19 




h r absent 


2 


2.4 


0.0 


390 


0.19 




total 


58 


0.72 


0.97 


520 


0.20 


hrl3X + + 


both absent 


7 


0.0 


0.0 


340 


0.21 




h f^ absent 


8 


0.0 


0.66 


360 


0.21 




It^r absent 


11 


0.75 


0.0 


380 


0.22 




total 


67 


0.73 


0.64 


480 


0.22 



Yields with small k and small burst size must have this effect, but are evi- 
dently not very important in the data of table 5, since the different classes are 
very similar with respect to k and burst size. In short, it is necessary to con- 
clude that there are at least two or three genetic exchanges per bacterium, 
independently of the mechanism by which recombinants arise. 

A different kind of estimate of the number of exchanges per bacterium is 
obtained from the number of recombinants actually found. The data are sum- 
marized in table 6 in the form of distributions of numbers of the several re- 
combinant types. One finds on the average 3.4 recombinants of any one kind 
per bacterium. This evidently furnishes an upper limit to the number of ex- 
changes per bacterium, insofar as exchanges yield viable and countable pro- 
geny. This result, taken in conjunction with the preceding estimates, leads to 
several remarkable conclusions. 

First, since the two methods of estimation, one minimal and one maximal, 
yield about the same result, there must be in fact only two or three exchanges 
per bacterium in the crosses between closely linked factors. 

Second, the recombinants must undergo little multiplication after they 
arise in the cell. 

Third, the conditions of viral growth in different bacterial cells must be 
equally favorable to genetic recombination; otherwise a larger proportion of 
bursts would fail to show recombinants. 



166 



60 A. D. HERSHEY AND RAQUEL ROTMAN 

These conclusions are substantially confirmed by the distributions of num- 
bers of recombinants shown in table 6, which are essentially of the Poisson 
type, with variance only moderately greater than the mean. In other words, 
the individual particles of any one recombinant must often arise independ- 
ently of each other in the same bacterial cell, and with equal probability in 
different bacterial cells. The deviations from the Poisson distribution are 
nevertheless significant, and can be attributed to a moderate amount of growth 
of recombinants. 

Table 6 

Dislrihutions of Numbers of Recombinants in Single Bursts from 
the Crosses hXrlJ and It rl3Xu>ild type 

The Poisson distributions show the numbers of tubes in the various classes expected if there 
were no growth of recombinants. The distribution with mean 3.4 is appropriate to the hypothesis 
of reciprocal exchange, and the distribution with mean 1.7 to the hypothesis of non-reciprocal 
genetic interaction. 



NO. OF RE- 
COMBINANTS 
PER TUBE 




NO. OF TUBES FOUND 




POISSON 
DISTRIBUTH 




h+r+ 


lir 


hr+ 


h+r 


ONS 





12 


4 


15 


18 


2.1 


11.5 


1 


11 


11 


11 


14 


7.2 


19.6 


2 


9 


8 


8 


6 


12.2 


16.7 


3 


6 


9 


11 


9 


13.8 


9.5 


4 


3 


7 


6 


4 


11.7 


4.0 


5 


7 


3 


4 


2 


8.0 


1.4 


6 


1 


2 


3 


6 


4.5 


0.4 


7 


3 


6 





2 


2.2 


0.1 


8 


2 


2 


3 


3 


0.9 


0.0 


9 


1 


2 


2 


1 


0.3 


0.0 


10-18 


3 


2 


3 


2 


0.1 


0.0 


19-27 





1 


1 





0.0 


0.0 


28-36 





1 








0.0 


0.0 


No. of tubes 


58 


58 


67 


67 


63 


63 


Mean per tube 


3.1 


4.5 


3.3 


2.8 


3.4 


1.7 


Variance 


9.2 


27 


16 


10 


3.4 


1.7 



The conclusion that genetic recombination is not suppressed in some of the 
bacteria is also supported by the results of the other crosses, in which no bursts 
yielding both parental viruses and lacking both recombinants were found. Only 
one burst, from the cross hXr7, failed to show one of the recombinants; it 
contained 95 percent of the h parent. 

INDEPENDENCE BETWEEN PROPORTION OF RECOMBINANTS AND BURST SIZE 

The data of table 4 do not show any significant correlation between burst 
size and proportion of recombinants, which means that the number of re- 
combinants must be very nearly proportional to the total yield of virus in 
single bursts. 

This conclusion must be qualified in view of the following considerations. 



167 



RECOMBINATION IN BACTERIOPHAGE 61 

Some of the tubes contain one or more mixed bursts plus one or more unmixed 
bursts. Such tubes would tend to show less than average values of k and less 
than average yields of recombinants together with greater than average burst 
size. The number of such tubes is about five percent of the total in our experi- 
ments and these tend to be concentrated in the class k^O.2. This probably 
explains the negative correlation between proportions of recombinants and 
burst size among tubes selected for small k. 

A larger proportion of the tubes, between 10 and 20 percent in our experi- 
ments, contain two or more mixed bursts. These tubes would tend to have 
greater than average k and greater than average burst size, but will not show 
exceptional proportions of recombinants. These tubes have the effect of weak- 
ening any correlation that may exist between burst size and proportion of re- 
combinants, especially among the class k^0.21. The data previously consid- 
ered do not, therefore, exclude the possibility of a weak positive correlation 
between proportion of recombinants and size of bursts. 

Advantage was taken of the finding that high pH reduces size of bursts, to 
examine the relation between burst size and yield of recombinants in another 

Table 7 

Small Bursts from the Cross hXr7 in Broth of pH 9.0 

See legend table 4. The proportions of recombinants have not been corrected for dispropor- 
tionate yields of parental types of virus. The individual cultures contain an average of about 1.4 
mixed bursts. 



NO. OF 
BURSTS 


k 


4k 


X 


y 


fi 


r(x, y) 


r(n, x-fy) 


20 
34 
54 


^0.20 

^0.21 

all 


0.61 
0.93 
0.81 


6.3 + 3.8 
7.4±3.5 
7.0±3.7 


10.7±6.2 
10.5 + 4.6 
10.6±5.2 


99 ±50 
123 + 73 
114±66 


0.53 
0.61 
0.56 


-0.34 
0.35 
0.12 



way. Since increased pH was found also to cause some of the bacteria to fail 
to liberate virus, the single burst technique was chosen. 

Bacteria were infected in the usual way with mutants h and r7 in broth of 
pH 6.8. At the end of the five minute adsorption period, and without treatment 
with antiserum, single burst cultures were prepared after diluting in broth of 
pH 9.0. Preliminary experiments showed that the yield of virus was complete 
under these conditions within one hour, and than no inactivation occurred 
during two additional hours at 37°C, or overnight in the refrigerator. Entire 
samples were plated on single mixed indicator plates. One successful experi- 
ment yielded 66 bursts, of which 54 were mixed, among 104 tubes receiving on 
the average 1.4 bacteria per tube. Evidently most of the bacteria liberated 
some virus. The average burst size, after subtracting the virus carried over 
from the input (totalling 36 particles per tube), was only 114 per tube, or about 
70 per bacterium corrected for the probable multiples. The average proportion 
of recombinants was nevertheless of normal size (table 7). 

It will be noticed also that the correlation between the numbers of the two 
recombinants in these bursts is exceptionally good. The correlation is, however. 



168 



62 A. D. HERSHEY AND RAQUEL ROTMAN 

slightly exaggerated owing to the fact that the 54 tubes contain on the average 
mixed yields of virus from about 1.4 bacteria per tube. 

THE EFFECT OF A SHORT PERIOD OF ADSORPTION AND LOW MULTIPLICITY 
OF INFECTION ON THE DISTRIBUTION OF RECOMBINANTS 

The following experiment shows that when the multiplicity of infection in 
the cross hy.r7 \s reduced from five to about one of each viral type per bac- 
terium, the distribution of recombinants among single mixed bursts is little if 
any altered. 

Six crosses were made in the usual way, except that the period allowed for 
adsorption was reduced to one minute, without reducing the total input of 
virus. The amount of virus adsorbed was too small to be measured, but the 
multiplicity of infection can be estimated from the data given below. 

The single burst cultures collected from the six crosses are sufficiently simi- 
lar to be considered together. The mean number of bacteria per tube for the 
six sets, determined by colony counts from the growth tubes immediately 
before adding virus, is 0.23. The mean number of infected bacteria per tube 
determined by plaque counts of samples taken before lysis is 0.20. The mean 
number of bursts per tube calculated from the proportion, 141 out of 720, of 
tubes containing virus is 0.22. The 141 tubes therefore contained about 157 
infected bacteria. 

From the distribution of viral types among the tubes, namely, 69 containing 
h only, 22 containing r only, 50 containing both, and 579 containing neither, 
one finds the probable distribution with respect to bacteria to be 80 infected 
with h only, 28 with r only, 49 both, and 46 neither. The multiplicity of in- 
fection is therefore about 1.0 with respect to h, and 0.48 with respect to r. One 
can estimate further that about seven of the tubes contained one or more 
mixed bursts plus one or more unmixed; and that about three contained both 
h and r bursts without any mixed bursts. Also, among the mixedly infected 
bacteria, 45 percent were infected with one particle only of each viral type. 

In making the above computations we have neglected the probability that 
the h mutant suppresses the growth of r in some bacteria adsorbing both 
types of virus. The apparent inequality of infection is probably due in some 
part to this effect. However, in other experiments with low multiplicity of in- 
fection with h and r7 designed to check this point, the split into A, r, and mixed 
yielders was nearly equal. It seems likely, therefore, that in the experiments 
reported here the two viral types were unequally adsorbed for unknown 
reasons. 

The tubes containing only one viral type may be dismissed by saying that 
their average content of virus did not differ significantly from that of the 
mixed yields, and that the h and r yields per bacterium were the same. There 
was one exceptional burst containing only h r+ and h r phenotypes. The char- 
acteristics of the remaining 49 cultures containing h and r virus are summarized 
in table 8. The data show no unusual features excepting the small burst size, 
which is a direct effect of the low multiplicity of infection, and the somewhat 



169 



RECOMBINATION IN BACTERIOPHAGE 63 

small proportion of recombinants, probably due to the appreciable number of 
superimposed unmixed bursts. 

Table 8 

Single Bursts from the Cross hXr7 with Low Midtiplicity of Infection 

See legend table 4. The proportion of recombinants have not been corrected for dispropor- 
tionate yields of parental types of virus. Adsorption time one minute. Multiplicity 1.0 h and 0.5 r 
per bacterium. 



NO. OF 
BURSTS 


k 


4k 


X 


y 


ii/lOO 


r(x, y) 


rCn, x+y) 


17 

32 
49 


^0.20 

^0.21 

all 


0.59 
0.95 
0.84 


5.3±4.0 
4.4 + 3.1 
4.7±3.5 


4.5±3.7 
6.1 + 5.4 
5.7 + 4.9 


2.9 + 1.6 
3.5 + 2.0 
3.2±1.9 


0.38 
0.20 
0.24 


-0.17 
-0.33 
-0.30 



The principal point to be made here is that the variation in yields of recom- 
binants from tube to tube is not exceptional. In this connection it must be 
mentioned that 5 of the 49 tubes contained h and r virus w^ithout any recom- 
binants. Of these, one was exceptional in containing only 24 viral particles, and 
another for the extreme disproportion of parental types (88 percent h). The 
remaining three, each containing from 450 to 570 particles, are probably 
superimposed unmixed h and r bursts. 

It should be noted also that the correlation between the yields of the two 
recombinants in this set is not significant, since the observed correlation is 
exaggerated by the tubes containing unmixed h and r bursts without recom- 
binants. Whether the poor correlation is accidental, or an effect of the low 
multiplicity of infection, remains to be determined. The negative correlation 
between burst size and proportion of recombinants can, however, be ascribed 
to the superimposed mixed and unmixed bursts, as well as to the unmixed 
h and r bursts, in some of the tubes. 

IDENTITY OF RECOMBINANTS WITH THE CORRESPONDING ANCESTRAL TYPES 

According to any simple hypothesis of factorial recombination, one expects 
the recombinant virus arising in crosses not to differ genetically from the cor- 
responding ancestral type. Two kinds of test indicate that this is so. In the 
first kind of test (Hershey and Rotman 1948) stocks of the phenotypic wild 
type arising from the cross between two different r mutants were backcrossed 
to authentic wild type. No r mutants appeared in such crosses, and it was 
concluded that the stocks were genetically identical. 

The second kind of test is the following. The double mutant // r7, itself ob- 
tained by crossing the two single mutants, was crossed with wild type and the 
recombinants h and r were re-isolated. These were then tested by making the 
homologous (parental h by recombinant h and parental r by recombinant r) 
and heterologous (parental h by recombinant r and parental r by recombinant 
h) back crosses. In both cases, the homologous cross yielded only one type of 



170 



64 A. D. HERSHEY AND RAQUEL ROTMAN 

virus, and the heterologous cross yielded recombinants in the same proportion 
as found with the parental stocks. These tests show not only that the recom- 
binants contain the same genetic markers as the corresponding progenitive 
types, but also that the region between the markers is unchanged. 

MIXED YIELDS CONTAINING ONLY ONE PARENTAL TYPE OF VIRUS 

Only four mixed bursts lacking one of the parental types of virus were found 
among the experiments reported in this paper. One, from the cross hXrl, 
contained 86 percent h^r and 14 percent h r. A second, from the cross hXr7 
contained 99.6 percent h r+ and 0.4 per cent h r. A third, from the same cross 
with low multiplicity of infection, contained 83 percent h r+, and 17 percent 
h r. The fourth, from the cross h r7X wild type, contained 23 percent h r+ and 
77 percent h r. In this case the yield of h r"*", which appeared to be homogenous, 
formed atypical plaques and proved on isolation to differ from any known 
mutant of T2H. It seems reasonable to suppose that these exceptional bursts 
contained progeny stemming from mutants contaminating the parental stocks 
of viius. On the other hand, the one exceptional burst following low multi- 
plicity of infection, together with the failure to find similar bursts among the 
crosses involving h and rl3, suggest that a different interpretation should be 
looked for. Genetic tests which might have clarified this point are lacking. 
For the present we conclude, as a first approximation, that recombinants arise 
only in those bacteria in which both parental types of virus succeed in multi- 
plying. 

It may be added here, because the question arises in connection with these 
exceptional bursts, that no correlation can be seen between the proportion 
in mixed bursts of the total virus containing the h allele, and the pioportion 
of the recombinant virus containing the h allele. We have therefore omitted 
this datum from the tables. 

DISCUSSION 

In collecting and analyzing the data just described, we have had in mind 
the following questions. Does genetic exchange occur in the course of matings 
between viral particles, or is it the expression of a mechanism of growth such 
as that visualized by Luria (1947), according to which the multiplying units 
in the cell are not phage particles, but simpler structures derived from them? 
Can the linkage relations represented in fig. 1 be interpreted in terms of linear 
chromosome-like structures? Are the genetic exchanges reciprocal, as one ex- 
pects for simple cases of crossing over, or must one look for an alternative 
mechanism more intimately connected with the mode of reproduction of 
the virus? 

It was soon apparent that the data for crosses between linked and unlinked 
factors tended to give different answers to these questions, and we were led 
to consider a model based on two distinct mechanisms of exchange. The neces- 
sity for this arises from the following facts. 

First, the linkage data indicate a limitation at about seven percent to the 
proportion of wild type found in crosses between linked factors, which is dif- 



171 



RECOMBINATION IN BACTERIOPHAGE 65 

ficult to reconcile, in terms of a single mechanism, with the existence of a 
second class of crosses yielding about 15 percent of wild type. 

Second, the correlation between proportions of the two recombinants in the 
cross hXr7, and the lack of a corresponding correlation in the cross hXrl, is 
incompatible with a single mechanism for the two crosses. It must be recalled, 
however, that the correlations found are too weak to be wholly convincing. 

Third, Luria's (1947) evidence for a mechanism of independent multiplica- 
tion and transfer of subunits of the virus, and ours for a system of linkage, re- 
quire dissimilar types of interpretation. 

The model to which these considerations seem to lead is described below, 
but we do not consider that we have decisive answers to any of the questions 
originally posed. The remainder of this discussion is of value only insofar as 
it clarifies the questions, and systematizes the experimental results so far ob- 
tained. 

The two linkage structures bearing the markers rl and h, respectively, are 
assumed to be examples of the class of independently multiplying subunits of 
the virus whose minimal number Luria and Dulbecco (1949) estimate at 
about 25. The reconstitution of virus from these units must be regulated in 
such a way that each particle receives one representative of each kind of unit. 
In the cross hXrl the choice between h and h+, and between r and r+, is de- 
cided nearly at random to yield on the average 37 percent of recombinants and 
63 percent of the parental types in bacteria yielding equal numbers of the two 
parents. The deficit of recombinants below 50 per cent is unexplained, but may 
be thought of as an effect of incomplete mixing between neighboring clones 
of multiplying virus in the cell. 

According to this hypothesis one expects from the cross hXrl proportionate 
yields of the two recombinants in a single burst to be: 

/ n(/^+) \/ n(r+) \ 

p(/^+r+ = m( I — (4) 

^^ \n(/^+) + n(/0/\n(r+) -f n(r)/ 

/ n(//.) \/ n(r) \ 
p(//r) = m( • — — — (5) 

where the expressions on the left refer to proportions of recombinant virus, 
the corresponding expressions on the right refer to the intra cellular yields of 
the respective unit linkage structures, and the coefficient m expresses the 
fraction of the intracellular virus which may be regarded as a random mixture 
of the two parental types, the remainder being considered unmixed. If the 
structures carrying the markers //, h^, r, and r+ grow independently in the 
cell, their yields will fluctuate independently, and no correlation will be ex- 
pected between the numbers of the two recombinants in sufficiently small 
yields of virus. This expectation is borne out by the data for viral yields from 
single bacteria. The average yields of the two recombinants are equal, however, 
showing that the several unit structures grow at equal rates. According to 
equations (4) and (5), the proportionate yield of recombinants should not be 



172 



66 A. D. HERSHEY AND RAQUEL ROTMAN 

influenced by burst size unless the latter has an effect on the fraction m. 
According to the data it does not, and m is a parameter having the average 
value 37/50. This interpretation requires that 26 percent of either kind of 
parental virus in the cell should multiply in effective isolation from the other 
parent. 

It will be noticed that the expressions (4) and (5) reduce to mk if one makes 
the approximations mentioned in connection with equation (3). This provides 
a theoretical basis for the correction we have applied to yields of recombinants 
in crosses between unlinked factors. An analogous justification for its use in 
crosses between linked factors will appear in the discussion to follow. 

The predictions for the cross //Xr7 are different from the preceding case, be- 
cause here the markers are situated on homologous linkage structures, so that 
recombination requires something like crossing over, which in turn requires 
something like synapsis. The expected fractional yields of recombinants are 

p(/i+r+) = p{hr) = I msc, (6) 

in which m has been defined previously, s is a fraction independent of m ex- 
pressing frequency of pairing, and c is a crossover frequency. In order to make 
c independent of m and s it is evidently sufficient to define the product ms as 
the fraction of the viral yield made up of particles in which the marked unit 
has descended from an unlike synapsed pair. The application of (6) leads to 
ambiguity if exchanges can occur between descendants of unlike synapsed 
pairs (Hershey and Rotman 1948). In what follows this difficulty does not 
appear to be very serious, but no rigorous analysis has yet been attempted. 

For the cross hXr7, if exchanges are reciprocal, one expects the correlation 
between proportions of the two recombinants to be disturbed only by fluctua- 
tions in relative growth of the two exchange products, in contrast to the cross 
hXrl, where the correlation is subject to fluctuations in the growth of four 
independent units. That these fluctuations are individually considerable is 
shown by the variations in relative and total viral yields in mixedly infected 
bacteria. A weak but probably significant correlation between proportions of 
the two recombinants is nevertheless visible in the cross hXr7. No such cor- 
relation can be seen in the cross hXrl. If the mechanism of exchange for these 
two crosses were the same, the greater correlation would be expected in the 
cross hXrl, which gives the larger yield of recombinants. 

The predictions for the cross hXrIJ are the same as for hXr7 except as 
modified by the much smaller yield of recombinants, presumably owing to a 
smaller frequency of crossing over. We have shown that in this cross the re- 
combinants come from two or three individual exchanges per bacterium, and 
that there is little growth of recombinants subsequent to exchange. These cir- 
cumstances ought to be favorable for testing the hypothesis of reciprocal ex- 
change. The data are nevertheless inconclusive of this point. 

The experiments provide information about the sequence of events in the 
cell. A mechanism of exchange limited to an initial phase of multiplication is 
ruled out by the following consideration. If exchange occurred at a time when 
there were few replicas in the cell, any cross yielding a small average number of 



173 



RECOMBINATION IN BACTERIOPHAGE 67 

recombinants would show some individual bursts containing no recombinants 
and others containing a very large proportion, especially at low multiplicity 
of infection. Instead one finds a comparatively uniform yield of recombinants, 
and the distribution of their proportions is not affected by the multiplicity of 
infection. 

Also if exchanges occurred freely throughout the period of multiplication of 
the virus, one would expect considerably greater variations in yields of recom- 
binants than we have found. For instance, in the cross hXrlJ, in which there 
are only two or three exchanges per bacterium, the variations in yields of re- 
combinants are not much greater than those expected to result from a random 
variation in the number of exchanges alone. Moreover, most of the bacteria 
yield only a few recombinants, so that little growth can have occurred sub- 
sequent to exchange. The conclusion is unavoidable that the exchanges are 
limited to the terminal phase of multiplication, or at any rate that recombi- 
nants are prevented from multiplying appreciably in most of the bacteria. 

It is remarkable that the variations in proportions of recombinants are so 
little dependent on the degree of linkage (as between hXr7 and hXrlJ), or on 
the postulated mechanism of exchange (as between the above and hXrl). The 
coefficients of variation in proportions of individual recombinants among 
single bursts are, for hXrl, about 40 percent; for hXr7, about 60 percent; and 
ioxhXrlS, about 100 percent. This circumstance also supports the inference 
that the exchanges are limited to a late phase of multiplication. 

The hypothesis stated permits one to examine further the structure of the 
linkage units. Since crosses between rl3 and any of the mutants belonging to 
the group closely linked to r7 yield about the same proportion, seven percent, 
of wild type (Hershey and Rotman 1948), it might be supposed that one 
crossover between the distant markers is always accompanied by several 
others, so that 50 percent of the progeny of synapsed pairs of the units r7 and 
rl3, for example, would be recombinant types. If this supposition is correct 
the terms of equation (6) can be evaluated by setting c = 0.5 and p(/i+r+) = 0.07. 
This gives 0.28 for the average fraction ms of virus descending from unlike 
synapsed pairs. If this fraction is assumed to be the same in other crosses in- 
volving the same linkage structure (the cross h r7Xrl3 reported in this paper 
suggests that it is), one can write for them 

p (wild type) = 0.14 c, (7) 

where c is the appropriate crossover frequency and the proportion of wild type 
is experimentally measured. The data for the three point crosses involving 
r2, r3, and r6 (Hershey and Rotman 1948) are examined from this point of 
view in table 9. The proportions of wild type have been calculated for random 
crossing over between unit linear structures, using the crossover frequencies 
given by (7). It will be seen that the data are entirely compatible with the 
hypothesis tested. Additional tests of this kind are needed, however. 

It will have been noticed that the average yield of recombinants in crosses 
between distant linked factors is very nearly half that found for unlinked 
factors. Dr. M. Delbruck has pointed out to us that this relationship can be 



174 



68 A. D. HERSHEY AND RAQUEL ROTMAN 

understood in terms of random pairing between homologous structures. In the 
simplest case one visualizes unrepeated pairing, that is, pairing limited to a 
phase in which there is no multiplication, and during which no structure finds 
more than one partner. For this case, the frequency of synapsis in equation 
(6) is simply the ratio of the number of unlike homologous pairs to the total 
number of homologous pairs. This ratio can be written 

2ab 
s = (8) 

(a + b)(a + b-l) 

where a and b are the respective numbers of the two unlike homologous 
structures. Inspection of (8) shows that this ratio is essentially 2k as given by 
equation (3) when a+b is large compared to unity. Instead of equation (6) we 
have, therefore, 

p (wild type) = mkc, (9) 

from which the proportion of either recombinant expected in crosses between 
distant linked factors can be computed as follows. 

The parameter m, taken to be the average fraction of virus randomly mixed 
in the cells at the time of reconstitution of virus from subunits, was found to 

Table 9 

Three Point Linkage Tests of Linear Structure 

The symbol r2,3 refers to the double mutant containing r alleles at the loci r2 and r3, etc. 
ci is the crossover frequency for the region between r2 and t3 . 
C2 is the crossover frequency for the region between rS and r6. 
The locus r3 is assumed to lie between r2 and r6. 
The factor 0.14 is explained in the discussion. 



CROSS 


FRACTIONAL YIELD OF WILD TYPE 
EXPECTED FOUND 


COMPUTATION 


r2Xr3 


— 


0.020 


ci = 0.020/0.14 = 0.14 


r3Xr6 


— 


0.014 


C2 = 0.014/0.14 = 0.10 


r2Xr6 


0.030 


0.024 


0.14[Ci(l-C2)+C2(l-Ci)] 


r2,3Xr6 


0.012 


0.008 


0.14c2(l-ci) 


r3,6Xr2 


0.018 


0.014 


0.14ci(l-C2) 


r2,6Xr3 


0.002 


0.003 


0.14 C1C2 



be 37/50 in crosses between unlinked factors. In (9) we require the corre- 
sponding fraction at the time of pairing, and assume this to be the same. The 
average of k for bacteria giving mixed viral yields (table 4) is 0.21, or 0.19 if 
one includes the ten percent of bacteria yielding only one type of virus. If k 
and m vary independently, their mean product is the same as the product of 
means, or 0.14 averaged over all bacteria. The average yield of either recom- 
binant from equation (9), for factors sufficiently far apart so that c = 0.5, 
is accordingly seven percent, computed solely from the data for crosses 
between unlinked factors. This is the maximum actually found in crosses 



175 



RECOMBINATIOxN IN BACTERIOPHAGE 69 

between linked factors (Hershey and Rotman 1948). Equation (9) also pre- 
dicts, in agreement with the data for single bursts, proportionality between 
yields of recombinants and k. 

The agreement supports the inferences previously drawn that the markers 
r7 and rl3 are attached to the same linkage unit, and that the frequency of 
crossing over between them is 0.5. It suggests that the pairing itself is complete 
without appreciable repetition, and occurs at random except that about 26 
percent of the units of each kind are effectively segregated from their opposite 
numbers. The measure of this segregation, m, is on this view the same for 
crosses between linked and unlinked factors. On the other hand, this inter- 
pretation cannot be rigorously correct, because one can show by multiple 
factor crosses (Hershey and Rotman 1948) that repeated exchanges, or ex- 
changes among three viral particles, occur. An estimate of the amount of 
repeated pairing has not yet been attempted, except that the considerations 
just offered suggest either that it is small, or that random pairing is limited to 
a small proportion of the population. 

It follows from equation (9) that the interpretation in terms of orderly 
pairing accords with the fact, otherwise very puzzling, that the proportion of 
recombinants is not affected by size of burst even in crosses between linked 
factors. 

It has been seen that the linkage data support fairly well the idea of linear 
structure, but independent evidence for crossing over is meagre. According 
to any simple model of reciprocal exchange, a correlation between proportions 
of sister recombinants in individual bursts would be expected. This expectation 
has been only partially realized, and the question arises whether the linkage 
data themselves require the crossover hypothesis. The following model, 
suggested by Dr. A. H. Sturtevant, shows that they do not, and also shows 
that the question of reciprocity is closely connected with the question whether 
the exchanges are material transfers. 

Suppose that the replication of linear structures occurs zipperwise along the 
pattern from one end to the other, but that the partners separate prematurely 
to yield fragmentary replicas. Additions to the fragments are subsequently 
possible only after pairing with the same or another homologous structure, 
which in mixedly infected bacteria could belong either to the same or a different 
parental line. Genetic recombination in a two factor cross will depend, then, on 
the contingency that the two marked regions of a given replica be laid down 
one after the other on homologous structures from the two unlike parents. 
With simple assumptions, all the consequences of the crossover hypothesis 
(equation (6)) follow from this model, except that the independent origin of 
the two recombinants provides an additional source of independent variation 
in their numbers. 

The complications peculiar to this model have to do principally with the 
evidence that exchanges occur only during the terminal phase of growth. 
These complications are not very serious if one assumes that during early 
stages of growth the probability is great that a fragment will be started and 
completed on patterns belonging to the same parental line; that is, that the 



176 



70 A. D. HERSHEY AND RAQUEL ROTMAN 

mixing of the cell contents is relatively incomplete, and the distance between 
unlike clones relatively great, for small total populations. It has to be stipu- 
lated further that the terminal mixing is independent of the final concentra- 
tion of virus in the cell, to account for the lack of dependence of proportion 
of recombinants on burst size. Some hypothesis of this sort may prove useful 
if further experiments fail to strengthen the present evidence for reciprocal 
exchange. 

It is notable that two very different lines of evidence, ours and that of 
LuRiA (1947), have led to the idea of independently multiplying subunits of 
the virus. Our results differ from Luria's only in calling for a system of linkage 
superimposed on the set of independent units. It remains to be seen whether 
a combination of genetic and radiological techniques bears out the present 
conclusions, and perhaps leads to an identification of the radiation-sensitive 
units with the linkage structures. 

SUMMARY 

Genetic recombination between two viruses differing by two mutational 
steps has been studied by infecting bacteria with the pair, and counting the 
numbers of the four types of virus found in yields from single bacteria. The 
crosses so examined include hy.rl (unlinked), AX^Z (linked), and hXrl3 
(closely linked), where h refers to a mutant of altered host range, and rl, r7, 
and rl3 are different mutations producing the same alteration in type ol 
plaque. The reverse crosses, ^rXwild type, were also studied. The results may 
be summarized as follows. 

Nearly all mixedly infected bacteria yield both parental types of virus and 
two recombinants, according to the scheme h-^r = h r-{--wi\d type. The ten 
percent or so of bacteria yielding only one of the parental types seldom or 
never yield any recombinants. The rest of the bacteria always yield two recom- 
binants, except for the occasional absence of one or both in the crosses between 
closely linked factors. 

The average yields of the two recombinants in any one cross are the same, 
and are independent of the direction of exchange, so that reverse crosses in- 
volving the same pair of mutant factors yield the same number of recombi- 
nants. The proportionate yields of recombinants from individual bacteria are 
independent of burst size, and of the total multiplicity of infection, but depend 
on the relative yields of the two parental types. The effect of the latter is not 
marked, however, and the variations from bacterium to bacterium must be 
chiefly the result of variations in the number of genetic exchanges and in the 
growth of recombinants subsequent to exchange. These variations may be de- 
scribed by saying that one finds a moderately skewed distribution, with mode 
less than the mean, and with mean and standard deviation dependent on the 
linkage relations as follows: for hXrl, 15 + 6, for hXr7, 7 + 4, for hXrlJ, 1 ± 1 , 
expressed in round numbers as percent of either recombinant in the total yield 
of virus. 

A weak but moderately convincing correlation between the proportionate 



177 



RECOMBINATION IN BACTERIOPHAGE 71 

yields of the two recombinants in individual bacteria is discernible in the cross 
hY^rl and its reverse, but not in the other crosses. 

In the cross hY.rl3 only two or three genetic exchanges occur during the 
multiplication of the virus in a single bacterial cell. These exchanges take 
place near the end of the period of multiplication of the virus. 

A hypothesis is outlined which is compatible with the genetic data and with 
the results of Luria concerning reactivation of irradiated virus in bacteria 
receiving two or more individually noninfective particles. The hypothesis is 
an extension of that of Luria. according to which one visualizes genetic 
interaction not between two viral particles, but between two sets of independ- 
ently multiplying chromosome-like structures. Genetic exchange occurs either 
by reassortment of these structures, or by something like crossing over between 
homologous pairs, depending on the structural relation between the genetic 
factors concerned. The interpretation made brings the linkage relations into 
superficial agreement with the requirements of linear structure, but there is 
little evidence that the genetic exchanges are reciprocal, and accordingly little 
evidence that they are material exchanges. 

LITERATURE CITED 

DelbrDck, M., 1945a Interference between bacterial viruses III. The mutual exclusion effect 

and the depressor effect. J. Bact. 50: 151-170. 

1945b The burst size distribution in the growth of bacterial viruses. J. Bact. 50: 131-135. 

1945c Effects of specific antisera on the growth of bacterial viruses. J. Bact. 50: 137-150. 
DELBRtJCK, M., and W. T. Bailey, Jr., 1946 Induced mutations in bacterial viruses. Cold Spring 

Harbor Symp. Quant. Biol. 11: 33-37. 
Delbrxjck, M., and S. E. Luria, 1942 Interference between bacterial viruses I. Interference 

between two bacterial viruses acting upon the same host, and the mechanism of virus growth. 

Arch. Biochem. 1: 111-141. 
DuLBECCO, R., 1949 The number of particles of bacteriophage T2 that can participate in intra- 
cellular growth. Genetics 34: (In press). 
Hershey, a. D., 1946a Mutation of bacteriophage with respect to type of plaque. Genetics 31 : 

620-640. 

1946b Spontaneous mutations in bacterial viruses. Cold Spring Harbor Symp. Quant. Biol. 

11:67-77. 
Hershey, A. D., and R. Rotman, 1948 Linkage among genes controlling inhibition of lysis in 

a bacterial virus. Proc. nat. Acad. Sci. 34: 89-96. 
Luria, S. E., 1947 Reactivation of irradiated bacteriophage by transfer of self-reproducing 

units. Proc. nat. Acad. Sci. 33 : 253-264. 
Luria, S. E., and R. Dulbecco, 1949 Genetic recombinations leading to production of active 

bacteriophage from ultraviolet inactivated bacteriophage particles. Genetics 34: (In press). 
Rider, Paul R., 1939 An introduction to modern statistical methods. ix-|-220 pp. New York: 

Wiley & Sons. 



178 



GENETIC RECOMBINATION AND HETEROZYGOSIS 
IN BACTERIOPHAGE 

A. D. Hershey and Martha Chase 

Department of Genetics, Carnegie Institution of Washington, 
Cold Spring Harbor, New York 

In this paper we summarize the principal features of inheritance in the bacterio- 
phage T2H, and describe some new experiments. 

The genetic structure of this virus has been analyzed in terms of mutational 
patterns (Hershey, 1946) and by recombination tests (Hershey and Rotman, 
1949). These two types of evidence agree in showing that mutational changes 
occur in localized regions of a complex genetic system. Mutations producing 
different effects usually occur at different loci, but one example of multiple 
allelism has been found (Hershey and Davidson, 1951). In this instance, the 
locus of the alternative mutations could be analyzed rather completely because 
most of the host-range mutations selected in a particular way proved to belong to 
a single allelic series. It was found that one pair of distinct mutants satisfied 
all three criteria of allelism listed below, and that another pair satisfied none of 
them. The criteria used were the following: 

(1) If the second of two successive mutations from wild type occur at the locus 
of the first, reversion to wild-type in a single step is possible. 

(2) No genetic recombination can be observed between allelic mutant pairs. 

(3) The map position of the locus is independent of its allelic state. 

BiPARENTAL RECOMBINATION 

The production of new genetic types of phage by intracellular interaction be- 
tween different bacteriophages was first observed by Delbruck and Bailey (1946), 
who mentioned genetic recombination as one of two possible interpretations of 
their result. The principle of genetic recombination was established by experi- 
ments with genetically defined stocks of the bacteriophage T2H (Hershey and 
Rotman, 1948, 1949). 

The main facts of genetic recombination in this bacterial virus can be illus- 
trated by examples of the interaction between two classes of mutant. Rapidly 
lysing (r) mutants are easily recognized by inspection of the plaques they produce 
on an agar plate seeded with sensitive bacteria. The plaques are larger, and have 
a sharper margin, than those of the wild-type. Host range {h) mutants are able to 
infect a suitable bacterial "indicator" strain that is resistant to the wild-type 

Reprinted by permission of the authors and the Long Island 

Biological Association from Cold Spring Harbor Symposia on 

Quantitative Biology, 16, 471-479 (1951). 

179 



.1. D. HERSHEY AND MARTHA CHASE 





FIG. 1. Progeny of the cross h X rl plated on mixed indicator. The large clear plaque 
is hr; small clear, h; large turbid, r; small turbid, wild-type. The eccentric clearings in the 
r plaque result from secondary h mutations. 

virus (Luria, 1945). The h mutants are normal with respect to type of plaque. 
By successive mutations, double (hr) mutants are readily obtained. The four 
kinds of virus, wild, h, r, and hr, can be recognized by plating on agar plates seed- 
ed with a mixture of sensitive (B) and indicator (B/2) strains of bacteria. Inde- 
pendently of the r character, h virus lyses both B and B/2 to produce clear 
plaques, and h^ virus lyses only B to produce turbid plaques. Independently of 
the h character, r virus produces large plaques, and r"*" virus produces small ones. 

Genetic recombination is observed in a "cross" in which sensitive bacteria are 
infected with a few particles of h and a few particles of r per cell. The viral proge- 
ny coming from the mixedly infected bacteria contains both parental types, to- 
gether with a certain proportion of the two recombinants (Fig. 1). Analogous re- 
combinants are found when two different r mutants are crossed. For example, 
the cross rl X r2 gives rise to the double mutant rlr2 and wild-type (Hershey and 
Rotman, 1948). 

In both types of cross, the yield of recombinants is characteristic for the 



180 



GENETICS OF BACTERIOPHAGE 



unit 
distance 

h ri3 



FIG. 2. Linkage relations among several genetic markers. 



mutant pair, and the numerical results can be summarized in the form of a genetic 
map, as shown in Figure 2. On this map we have shown only a few markers that 
we wish to refer to in this paper. The letter m (minute) stands for a small plaque 
mutant. It is well established that the three loci, rl , h,tand m assort independent- 
ly of each other, and that the loci linked to h are arranged in Hnear order. Dr. N. 
Visconti (personal communication) has recently confirmed the earlier results on 
these points by a new method of three-point testing. 

Triparental Recombination 

Important information about genetic recombination comes from experiments 
in which the frequency of triparental recombination is measured. If bacteria are 
infected with the three mutants h, m, and rl , the recombinant hmr can arise only 
by interactions involving all three. The results for this triple infection and its re- 
verse, hr X hm X mr, are shown in Table 1. About three per cent of the viral 
yield consists of the triparental recombinant in these crosses. This shows that in- 
teractions among three particles of virus occur with high frequency. A measure of 
this frequency can be expressed in terms of some artificial assumptions. If we 
suppose that multiplication precedes recombination, and that recombinants are 
formed during random successive pairings between phage particles, each particle 
would have to pair with three other particles to explain the results shown in Table 
1. The genetic factors used in these experiments are unlinked. The data for tri- 
parental recombination in experiments with linked markers (Hershey and 
Rotman, 1948) lead to the same quantitative conclusion. 

This conclusion is a stumbling block to the further understanding of the mecha- 
nism of genetic recombination, since it is extremely difficult to distinguish be- 
tween successive interactions by pairs, and other types of interaction that might 
involve larger groups. The available information is insufficient, therefore, to de- 
cide between alternative hypotheses of viral interaction so far considered. 



181 



.1. D. HERSHEY AND MARTHA CHASE 




.m^ 






FIG. 3. Mottled plaques from bacteria infected with r and r+ virus. 

Experiments with Clumped Phage 

The following observations are pertinent to the proper interpretation of experi- 
ments to be described in a later part of this paper. Since they are also of general 
interest, we record them separately. 

The r mutants are unique among the known mutants of phage in that mixed 
colonies of r and r"*" phage are easily recognizable as mottled plaques. Mottled 
plaques are conveniently prepared by infecting bacteria with r and r"*" phage, and 
plating before lysis so that the plaques originate not from single phage particles, 
but from the mixed population liberated locally when the bacterium lyses (Fig. 3). 
We have made use of this characteristic mottling as a test for the aggregation of 
phage particles. 

Platings of mixtures of r and r"^ phage do not show mottled plaques (except 
rarely by the overlapping of plaques), even after the mixture has been packed in a 
centrifuge to allow every opportunity for the particles to stick together. Mottling 
clumps can be produced, however, by agglutinating the mixture with antiserum. 

To prepare mottling clumps of phage, one adds to a mixture containing half r 
and half r"^ phage, at a concentration of about 10" particles per ml, an amount of 



182 



GENETICS OF BACTERIOPHAGE 

TABLE 1. TRIPARENTAL RECOMBINATION TESTS 

WITH THREE INDEPENDENTLY ASSORTING 

GENETIC FACTORS 



Cross 






Per cen 


it disi 


tribution 


in yield 












wild 


h 


r 


m 


hr 


hm 


mr 


hmr 


h X m X rl 


(1) 


25 


17 


22 


12 


9 


5 


7 


2 


hm X mrl X hrl 


(2) 
(1) 


25 
3 


18 
6 


15 
5 


20 
10 


4 
17 


10 
19 


5 
14 


3 

26 




(2) 


3 


9 


4 


9 


14 


26 


15 


20 



Results are shown for two independent experiments of each kind. 



antiphage serum just sufficient to reduce the titer of the mixture by a factor 2 to 
4 after equihbrium has been reached. This requires several hours at room temper- 
ature. The mixture now yields, on plating, about two per cent of mottled 
plaques. 

Mottled plaques coming from clumps containing the doubly marked phages hrl 
and wild-type are found, on sampling, to contain the two parental types of phage, 
together with the expected recombinants. The same is true of mottled plaques 
originating from bacteria infected with hrl and wild-type. This shows that there 
is no strong selection among the several genetic types of phage during plaque for- 
mation. It shows, incidentally, that the genetic result of a mixed infection is of 
the same kind whether the parental phage particles attach to the bacterium at the 
same pomt, as with mottling clumps, or at different points, as m the usual mixed 
infection. 

The mottling clumps are identifiable as clumps by their abnormal sensitivity to 
antiserum. When 90 per cent of a population containiiig two per cent of mottling 
clumps is inactivated by exposure (at low concentration, to avoid further aggre- 
gation) to antiserum, the proportion of mottling clumps among the survivors is 
less than 0.2 per cent. 

When a population of phage containing two per cent of mottling clumps (hrl + 
wild-type) is adsorbed to B so as to infect one bacterium in ten, and the infected 
bacteria are plated on B after inactivating the unadsorbed phage with antiserum, 
two per cent of the plaques from infected bacteria are mottled. The resistance of 
the adsorbed clumps to antiserum shows that both members of a mottling clump 
infect the same bacterium. When the same population is adsorbed to B/2, and 
the infected cells are treated with antiserum or merely washed to remove unad- 
sorbed phage, only hr plaques are formed on plates seeded with B. Evidently 
both members of a mottling clump have to make specific attachments to the bac- 
terium in order to produce a mixed yield. This result probably explains why we 
are unable to prepare populations contaming more than a small fraction of mot- 
tling clumps by agglutination with antiserum. Only those clumps in which the 
two kinds of virus are oriented in a manner favorable to the attachment of both 
to the same bacterium can produce mottled plaques. 



183 



.4. D. HERSHEY AND MARTHA CHASE 

Heterozygosis 

Mixed yields of T2H from bacteria infected with r and r+ phage always contain 
about two per cent of particles that give rise to mottled plaques. Samples of 
phage from these mottled plaques consist of approximately equal numbers of 
typical r and r"*" particles, with only traces of mottling phage particles. This dis- 
tinguishes motthng phage particles from phage particles containing an unstable 
genetic factor affecting the r character (Hershey, 1946). These produce sectored 
rather than mottled plaques, and the sectored plaques contain sectoring particles 
and r particles, but no r+. 

The mottling phage particles do not consist of clumps. This is shown both by 
the genetic data to be described presently, and more generally by the following 
consideration of mactivation data. 

Suppose, for example, that a mottling phage particle really consists of a small 
clump of j normal r+ and k normal r particles. If the population containing this 
clump is heated sufficiently to reduce the titer by a factor 100, the chance that the 
mottling property of the specified clump will survive is at most (1 — 0.99') (1 — 
0.99*=), which is very much less than the chance (0.01) that a given phage particle 
will survive. The proportion of mottled plaques originating from clumps in a small 
surviving fraction of the population will, therefore, be very much less than the 
proportion of mottling clumps in the original population. Mottling particles do 
not behave in this way. In experiments in which populations containing mot- 
tling phage were inactivated by heating, by antiserum, by ultraviolet light, or by 
beta rays from P^^ no appreciable decrease in the proportion of mottling phage 
among survivors was seen. (The populations were examined at ten per cent and 
one per cent levels of survival.) The mottling particles are inactivated as a unit 
by the agencies mentioned, and possess the same resistance to inactivation as the 
non-motthng phage particles in the population. 

We conclude that mottling phage coming from bacteria infected with r and r+ 
virus contains both parental markers in a particle of otherwise normal properties. 
Since there is adequate reason to call these markers allelic genes, the mottling 
particles are appropriately termed heterozygotes. We use this word without in- 
tending to imply any specific structural basis for the observed properties of the 
particles. 

There is no evidence that the heterozygotes can multiply in the heterozygous 
condition. The small proportion (roughly two per cent) of mottling phage par- 
ticles that is found in the mottled plaques originating from heterozygotes is also 
found in mottled plaques coming from bacteria infected with mixtures of r and r+ 
phage. The similarity of the two proportions suggests that the mottling particles 
have been formed in both instances during the development of the plaque. 
The proportion of r — r+ heterozygotes produced in bacteria infected with 
mixtures of r and r+ virus does not vary significantly from two per cent for five 
different r markers (Table 4, left haK), and is also about the same in crosses be- 
tween h and r mutants, to which we now turn. 

Information about the structure of the heterozygotes can be sought by analyz- 
ing them in terms of the segregants they yield. For this purpose, the viruses giv- 



184 



I 



GENETICS OF BACTERIOPHAGE 

ing rise to the heterozygotes must differ by at least two genetic factors. Our ex- 
periments have been hmited to differently linked pairs of h and r markers. The 
experimental method is simple. One samples mottled plaques containmg the 
segregants from heterozygotes and replates the phage progeny on mixed indi- 
cator. The types of virus recognized in this way form the basis for classifying the 
original heterozygote. 

By selecting mottled placjues, we limit the exammation to heterozygotes segre- 
gating for r and r+. The analysis yields information only about the pattern of 
segregation of the additional marker within this class. For any cross hr X wild- 
type, only three results are found. 

TABLE 2. PER CENT DISTRIBUTION OF r-r+ HETEROZYGOTES WITH RESPECT 
TO SEGREGATION PATTERN: CROSSES WITH EQUAL MULTIPLICITY OF 

INFECTION 



Segregants 


hrl 


X wild 


hr? 


X Wi 


ild 


h Xr7 


hrl 3 


X wild 


found 




(40) 




(20) 




(20) 




(2) 


hr-wild 




6 




6 









74 


h-hr 




49 




44 




55 




15 


r-wild 




45 




50 




40 




11 


h-r 













5 







No. tested 




253 




150 




129 




494 



Per cent yields of recombinants in each cross are shown in parentheses. 

(1) Segregants h and hr, corresponding to one parent and one recombinant of 
the original cross. Heterozygotes of this class lack the h^ marker, or lose it during 
segregation. 

(2) Segregants r and wild-type, corresponding to the second parent and second 
recombinant of the original cross. Heterozygotes of this class lack the h marker. 

(3) Segregants hr and wild-type, corresponding to the two parents of the origi- 
nal cross, coming from doubly heterozygous phage particles. These segregants 
are necessarily accompanied by their recombinants. 

Two qualitative results are evident. First, the heterozygotes segregate into 
pairs. This shows that segregation precedes multiphcation, or that daughter 
heterozygotes segregate in only one way. Second, the double heterozygotes do 
not segregate into the two recombinants, h and r, but only into the two parents, 
hr and wild-type. It should be noted, however, that these two alternatives could 
not be distinguished in the cross hrl X wild-type, because of the large yields of re- 
combinants in either case. 

The quantitative results are summarized in Table 2. The doubly heterozygous 
class is a small minority in the crosses involving unlinked or distant markers, and 
forms a surprisingly small majority even for the closely linked factors h and rl8. 

The inferences that can be drawn from Table 2 are supported also by tests on a 
smaller scale of heterozygotes from the crosses h X rl and h X rl3. 

In view of the small proportion of doubly heterozygous particles produced in 
the crosses hrl X wild-type, and hr7 X wild-type, there was some question 



185 



.4. D. HERSHEY AND MARTHA CHASE 

whether we were really measuring this proportion. Two possible sources of error, 
namely, accidental overlaps of two plaques, and clumps containing two or more 
phage particles, were excluded by the following experiment. Plates showing not 
more than 50 plaques (as opposed to about 100 in other experiments) were pre- 
pared from a population of phage from the cross hrl X wild-type, of which 90 per 
cent had first been neutralized by antiserum. The antiserum treatment should 
have eliminated any clumps of phage, and the small number of plaques per plate 
should have eliminated overlaps. Eighty-three mottled plaques were sampled 
from these plates, of which five proved to contain both parental types of phage. 
This is the same proportion foimd in 170 plaques examined in other experiments. 
The estimate of six per cent shown in Table 2 is therefore correct. 

The data of Table 2, together with the estimated total frequency (2%) of r-r+ 
heterozygotes, measure the frequencies of three classes of heterozygotes among 
the progeny of crosses between h and r markers. These frequencies are, for the 
cross hr7 X wild-type, 0.12 per cent hr-wild; 0.94 per cent h-hr; and 0.94 per cent 
r-wild; expressed in round numbers. The corresponding frequencies for the cross 
hrl 3 X wild-type are 1.48, 0.26, and 0.26 per cent, respectively. What other 
heterozygotes might we expect to find among these progeny? 

Two possible classes remam to be looked for; namely, the classes segregating 
into the pairs hr, r; and h, wild; coming from particles heterozygous for h but not 
for r. If the distribution of heterozygotes is symmetrical, that is, if the total 
frequency of heterozygosis for h is two per cent, and if the two undetected classes 
are of equal size, their individual frequencies would be 0.94 per cent for the cross 
hr7 X wild-type, and 0.26 per cent for the cross hrlS X wild-type. 

One of the midetected classes can be efficiently measured by sampling clear r+ 
plaques from platings of the progeny of the cross on mixed indicator, and retest- 

TABLE 3. THE FREQUENCY OF h-wild HETEROZYGOTES PRODUCED IN 
CROSSES BETWEEN hr AND WILD-TYPE 





hr7 X ivild 


hrl 3 X wild 


Per cent clear r+ plaques 


13 


1.3 


No. mixed/no. tested 


14/110 


33/112 


Per cent h-wild heterozygotes among progeny 






Expected 


0.94 


0.26 


Found 


1.7 


0.38 



ing to determine how many of the samples contain mixtures of h and wild-type 
phage, and how many contam h only. The proportion of clear r+ plaques that 
yield mixtures, multiplied by the proportion of clear r+ plaque-formers among the 
progeny of the cross, gives the frequency of h-wild heterozygotes in the popu- 
lation. 

The results of this measurement for two crosses are compared with the expecta- 
tion for symmetrical distributions of heterozygotes in Table 3. The findings are 
similar to those already described for r heterozygotes, namely: 



186 



GENETICS OF BACTERIOPHAGE 

In the cross hr7 X wild (distantly linked markers) the great majority of parti- 
cles heterozygous for h are not heterozygous for r. 

In the cross hrl3 X wild (closely linked markers) the singly heterozygous class 
is smaller, but the proportion of heterozygotes yielding recombinants is large 
compared to the proportion of recombinants among the progeny as a whole. 

In both crosses the total frequency of h-h~^ heterozygotes is roughly two per 
cent. 

We conclude that the formation of particles heterozygous for h and for r obeys 
identical rules. 

A method of somewhat similar principle is applicable to crosses between 
pairs of r mutants. This is important in that effects of linkage can be 
tested in different regions of the genetic map; unfortunately, not with very 
great precision. The principle of the test can be illustrated by the example 
rS X r4. Heterozygotes resulting from this cross should segregate to yield the 
following pairs; r3 + rSr4, r4 -f r2r4-, r2 + r4, r2 + wild, r4 + wild. This list 
includes all possible pairs excepting the recombinant pair, which is assumed to be 
absent. The first two and the last two classes will all be of the same size for 
reasons of symmetry, and the sum of the frequencies of the last three classes will 
amount to two per cent of the population. If the two factors are not linked, heter- 
ozygotes belonging to the last three classes, and only these, will yield mottled 
plaques. If the factors are linked, only the last two classes will yield mottled 
plaques, and the difference between the proportion found and two per cent will 
measure the size of the doubly heterozygous class yielding r2 -\- r4- The assump- 
tion that particles doubly heterozygous for linked factors do not produce mottled 

TABLE 4. YIELDS OF r+ AND MOTTLING PHAGE IN r CROSSES 





Per cent 


Per cent 




Per cent 


Per cent 


Cross 


r+ 


mottled 


Cross 




r+ 




mottled 


r2 X wild* 


41 


1.64 ± .24 


r2 X r4 


0.85 ± 


.17 


0.88 ± .17 




37 


1.69 ± .18 




0.80 ± 


.13 


0.70 ± .14 


rif. X wild 


41 


1.98 ± .10 


r2 X r7 


3.8 


± 


.33 


1.40 ± .18 










3.1 


± 


.37 


1.19 ± .17 


r7 X wild 


46 


1.63 ± .19 














40 


1.59 ± .21 


rU X rl* 


3.7 


± 


.32 


1.25 ± .19 










3.5 


± 


.29 


1.90 ± .29 


rl3 X wild* 


50 


1.88 ± .25 














54 


1.65 ± .33 


rl3 X r7* 


8.1 


± 


.50 


1.43 ± .17 










7.3 


it 


.52 


1.27 ± .17 


rl X wild* 


45 


2.17 ± .26 














47 


2.43 ± .21 


rl X r7 


17.9 


± 


1.2 


2.04 ± .26 








rl X rl3 


16.7 


± 


.66 


2.00 ± .27 



The results shown are means and their standard deviations computed from counts of about 
4000 plaques on 20 plates. The duplicate counts of the crosses marked with an asterisk were 
made from the same population on different days; the other duplicate counts represent in- 
dependent experiments. The per cent r+ in the left half of the table measures the equality of 
infection with the two parental viruses; in the right half it msasures the linkage between r loci. 



187 



A. D. HERSHEY AND MARTHA CHASE 

plaques is tested by examining plaques originating from bacteria mixedly infected 
with r mutant pairs, which shows that for map distances up to and including 10 
units (r? X rl3) the mottling is negligible. 

The results of tests of this kind with six pairs of r mutants are shown m Table 4. 
Three facts emerge. The frequency of heterozygosis with respect to five different 
r loci is at least approximately the same. The effect of close linkage between the 
loci r3 and r4 is to produce a sufficiently large class of doubly heterozygous phage 
particles to cause a sharp decrease in the yield of mottling phage. No significant 
effect is seen for crosses between markers separated by a distance of three or four 
units or more. The method is evidently valid in principle, but too inaccurate to 
yield detailed information. 

We return once more to crosses between phages carrying h and r markers to test 
the effect of unequal multiplicity on the pattern of segregation of the resulting 
heterozygotes (Table 5). The effect seen is to increase markedly the frequency of 

TABLE 5 PER CENT DISTRIBUTIONS OF r-r^ HETEROZYGOTES WITH RESPECT 
TO SEGREGATION PATTERN: CROSSES WITH 5-FOLD EXCESS OF hr OVER 

WILD-TYPE 

hrl X wild hr7 X wild hrlS X wild 

(40) (20) (2) 



Segregants found 



•,j 2 4 71 

hr-wild ^ 

h-hr 78 71 7 

r-wUd 20 25 12 

hr ..0 

No. tested 117 55 161_ 



Per cent yields of recombinants for each cross are shown in parentheses. 
the parent-recombinant class contaming the parental virus available in excess. 
This effect is not visible when the markers involved are closely linked. 

Another experiment yields information about the sequence of events in the cell. 
It is based on the work of Doermann (1948a), who has shown that infected bac- 
teria artificially lysed at various times during the latent period of viral growth do 
not yield any virus during the first half of the latent period, and that the yields 
rise linearly from zero to a maximum during the second half of the latent period. 
The current interpretation of this result is that the first half of the latent period is 
devoted to the multiplication of non-infective virus, and the second half to the 
conversion of non-infective into infective virus (Doermann and Dissosway, 1949; 
Luria, 1950; Hershey, 1951). During the second half of the latent period, the 
partial yields of virus can be obtained simply by adding cyanide to the cultures 
Doermann (1948b) has also shown that very large yields of virus can be obtained 
when lysis is delayed for several hours, as happens in cultures contammg high 
concentrations of bacteria infected with r+ virus. 

We have compared viral yields obtained from samples of the same bacterial 
suspension, infected with hr7 and wild-type, by adding cyanide ten minutes after 
infection, by spontaneous lysis at the end of the normal latent period (21 to 40 
minutes), and by spontaneous lysis delayed for five to six hours. The yields of 



188 



GENETICS OF BACTERIOPHAGE 

virus per cell were respectively 10, 250, and 1710. The yields of recombinants 
were respectively 17, 29, and 42 per cent of the total virus. The yields of r-r+ 
heterozygotes, however, did not differ significantly from two per cent in any of 
the three populations. The proportions of the different segregating classes were 
also the same among the heterozygotes in small samples from the first two yields. 
This experiment shows that the interactions between phage particles giving rise 
to recombinants (Doermann and Dissosway, 1949) and heterozygotes, are well 
under way by the time infective virus begins to form in the cell. The rise in pro- 
portion of recombinants during the latter half of the latent period, while the pro- 
portion of heterozygotes remains constant, suggests that the formation and segre- 
gation of heterozygotes may be continuing during this time. 

Summary of New Facts 

(1) When bacteria are infected with three kinds of phage carrying unlinked 
genetic markers, about three per cent of the progeny carry markers derived from 
all three parents. 

(2) When bacteria are infected with two phages carrying allelic markers, about 
two per cent of the progeny particles segregate during further growth to yield 
both kinds of phage. This is true for five different r markers, and an h marker. 
The particles segregating to yield r and r+ phage are conveniently studied because 
they produce mottled plaques. 

(3) The mottle producers are not clumps of particles because they are in- 
activated as single units of normal sensitivity by antiserum, heat, /3-rays, and 
ultraviolet light. Artificially prepared clumps are inactivated as multiple units. 

(4) There is no indication that the mottling particles can multiply before segre- 
gating. In view of facts (3) and (4), the segregating particles are called hetero- 
zygotes for the specified marker. 

(5) When the parental phage particles are marked at both h and r loci, the pat- 
tern of segregation shows the following characteristics: 

(a) About two per cent of the progeny are heterozygous for h, and about two 
per cent for r. 

b) The particles heterozygous for a single marker form four classes of approxi- 
mately equal size, segregating into h, hr; h, wild; r, hr; and r, wild; respectively, 
respectively. Thus single heterozygotes yield one parent and one recombinant 
with respect to the original cross. 

(c) When the two markers are linked, the double heterozygotes segregate to 
yield the two parents of the original cross, never the two recombinants, and never 
more than two kinds of phage. When the two markers are unlinked, these al- 
ternatives cannot be distinguished. 

(d) When the two markers are unlinked or distant, heterozygosis for one 
marker is almost independent of heterozygosis for the second, and the doubly 
heterozygous class amounts to only three per cent of the total number of hetero- 
zygotes. Thus in the crosses hr7 X wild-type and hrl X wild-type, which yield 
respectively 20 and 40 per cent of recombinants, the pooled heterozygotes segre- 
gate to yield about 48 per cent of recombinants. 



189 



.4. D. HERSHEY AND MARTHA CHASE 

(e) In the cross hrlS X wild-type (closely linked markers) the doubly hetero- 
zygous class makes up about 59 per cent of the total number of heterozygotes. 
This cross yields about two per cent of recombinants, but the pooled hetero- 
zygotes segregate to yield about 20 per cent of recombinants. 

(f) If the crosses involving unlinked or distant factors are varied by introduc- 
ing a 5-fold excess of one parent, the effect is to increase the frequency of the 
single heterozygotes segregating to yield that parent. This effect is not visible if 
the markers are closely linked. 

(6) The frequency of heterozygosis is independent of the yield of virus per bac- 
terium when this is decreased by premature lysis with cyanide, or increased under 
conditions of lysis-inhibition. 

Discussion 

Information about inheritance in bacteriophage T2H comes from the analysis 
of mutations and from recombination tests. These two techniques agree in show- 
ing that mutations occur in localized genes. Recombination tests reveal that the 
genes are organized into linkage groups. For one of these groups, it appears that 
the arrangement of genes is linear. Inheritance in bacteriophage is therefore 
amenable to the same kind of genetic analysis that has served to elucidate nuclear 
organization in other organisms. The limitations peculiar to viral genetics should 
not be overlooked. It is not possible to recover the immediate products of re- 
combination, unless the heterozygotes prove to be such; the mechanism of re- 
combination is unknown ; and cytogenetic techniques are inapplicable. 

The analysis of heterozygotes raises new questions about the mechanism of 
genetic recombination. The surprising result is that the great majority of the 
heterozygotes recovered from a two-factor cross segregate as if they were 
homozygous or hemizygous for one of the marked genes, unless these are very 
closely linked. This means that the heterozygotes found, which should perhaps 
be called residual heterozygotes, may not be representative of the heterozygotes 
formed in the cell. 

One feature of the residual heterozygotes is reassuring in this respect. The 
total frequency of heterozygosis is the same for five different r markers and one 
h marker. This makes it unlikely that the formation of residual heterozygotes is 
contingent on structural differences between different mutants. 

The questions raised by the peculiar segregation pattern of heterozygotes are 
clarified somewhat in terms of the following alternatives. 

(1) Residual heterozygotes may not be diploid particles, but particles con- 
taining one or more small extra pieces of genetic material. Double heterozygotes 
for distant markers contain two or more pieces. These pieces are substituted for 
the homologous pieces in one of the very early progeny of the segregating 
heterozygote. The residual heterozygotes need not differ from intracellular 
heterozygotes, and their production need not involve zygote formation. 

2) Residual heterozygotes may be formed preferentially from zygotes in 
which recombination has occurred, and receive one parental and one recombinant 



190 



GENETICS OF BACTERIOPHAGE 

set of genes. In this case the residual heterozygotes are diploid, but are not 
representative of the zygotes from which they come. 

(3) Residual heterozygotes may be representative zygotes that are doubly 
heterozygous in structure, but which undergo segregations accompanied by 
frequent losses to yield parental and recombijiant pairs. 

The third alternative can be excluded. In crosses involving the markers h 
and rl3, one finds only two per cent of recombinants among the whole progeny, 
and about 20 per cent of recombinants among the segregants of heterozygous 
progeny. To explain this in terms of alternative (3), one would have to assume a 
low frequency of intracellular zygote formation. This assumption is incompatible 
with the high frequency of triparental recombination observed. 

The questions about structure of heterozygotes can be generalized in the 
following way. We find that about two per cent of the progeny of crosses are 
heterozygous for each marker, and that the particles heterozygous for one are 
mostly not the particles heterozygous for the other, excepting close linkage. Since 
the frequencies are not specific for individual mutants, they are presumably 
independent of local structure, and every phage particle must carry doublings at 
one or more unmarked loci if total map distances are large. The alternatives 
(1) and (2) are to this extent applicable to all the progeny, and take the simple 
form: are phage particles diploid or not? 

It is reasonable to assume that the formation of heterozygotes and the for- 
mation of recombinants are related processes, but there is no evidence that 
recombinants have their primary origin in structures resembling the residual 
heterozygotes. Instead, recombinants and residual heterozygotes may be 
alternative products of other structures about which we have no direct mfor- 
mation. The residual heterozygotes have one characteristic that is suggestive in 
this connection : they segregate to yield one recombinant per heterozygote. The 
recombinants that are produced in crosses also have to be assumed to come from 
structures yielding one recombinant, to explain the independent or nearly 
independent distributions of sister recombinants among single cell yields of virus 
(Hershey and Rotman, 1949). 

The frequency of double heterozygotes provides a measure of linkage that is 
independent of the results of recombination tests. Both measures show that h is 
linked to rl3 and that r2 is linked to r4. The new measure is insensitive for large 
map distances since the crosses involving h and r7, and h and rl, which yield 
respectively 20 and 40 per cent of recombinants, produce the same number of 
double heterozygotes. 

Conclusion 

A preliminary analysis of heterozygous particles of the bacteriophage T2H 
raises new questions about the mechanism of genetic recombination, and suggests 
that new ideas are needed to explain this phenomenon. 

The work reported in this paper was aided by a grant from the Division of 
Research Grants and Fellowships, U. S. Public Health Service. 



191 



A. D. HERSHEY AND MARTHA CHASE 



References 

Delbruck, M., and Bailey, W. T., Jr., 1946, Induced mutations in bacterial viruses. 

Cold Spr. Harb. Symposium Quant. Biol. 11: 33-37. 
DoERMANN, A. H., 1948a, Intracellular growth of bacteriophage. Yearb. Carneg. Instn. 

47: 176-182. 
1948b, Lysis and lysis inhibition with Escherichia coli bacteriophage. J. Bact. 55: 

257-276. 
DoERMANN, A. H., and Dissosway, C. F.-R., 1949, Intracellular growth and genetics of 

bacteriophage. Yearb. Carneg. Instn. 48: 170-176. 
Hershey, a. D., 1946, Spontaneous mutations in bacterial viruses. Cold Spr. Harb. 

Symposium Quant. Biol. 11: 67-77. 
1951, Reproduction of bacteriophage. Vllth International Congress for Cell Biology, 

in press. 
Hershey, A. D., and Davidson, H., 1951, Allelic and non-allelic genes controlling host 

specificity in a bacteriophage. Genetics 36: 667-675. 
Hershey, A. D., and Rotman, R., 1948, Linkage among genes controlling inhibition of 

lysis in a bacterial virus. Proc. Nat. Acad. Sci., Wash. 34: 253-264. 
1949. Genetic recombination between host range and plaque type mutants of bacterio- 
phage in single bacterial cells. Genetics 34: 44-71. 
LuRiA, S. E., 1945, Mutations of bacterial viruses affecting their host range. Genetics 30: 

84-99. 
■ 1950, Bacteriophage: an essay on virus reproduction. Science 111: 507-511. 

DISCUSSION 

ViscoNTi (in reply to a comment by Horowitz) : The excess of parental types 
in Dr. Hershey's experiments can be eliminated by selecting a class of recombi- 
nants and scoring inside this class for a third character. Making use of the three 
markers, rl , h and m, the following cross was made: rl h rri^ X rl'^ h^ m. The 
yield w^as plated on B, so that no difference could be detected between h and /i"*". 
Of 1003 plaques observed, 177 were r+ m+, thus giving a recombination value of 
18 per cent. 128 of such plaques were "fished" and tested by a streaking method 
on B/2. Of the 128 tested, 63 were h and 65 h^. In another experiment, 92 
plaques were "fished" and tested by plating a sample on double indicator. 43 
were h; 47 were /i+; and two were mixed. The two mixed plaques account for the 
2 per cent of heterozygotes for the locus h. 



192 



THE STRUCTURE OF DNA 

J. D. Watson^ and F. H. C. Crick 

Cavendish Laboratory, Cambridge, England 
{Contribution to the Discussion of Provirus.) 

It would be superfluous at a Symposium on Viruses to introduce a paper on the 
structure of DNA with a discussion on its importance to the problem of virus re- 
production. Instead we shall not only assume that DNA is important, but in ad- 
dition that it is the carrier of the genetic specificity of the virus (for argument, see 
Hershey, this volume) and thus must possess in some sense the capacity for exact 
self -duplication. In this paper we shall describe a structure for DNA which sug- 
gests a mechanism for its self-duplication and allows us to propose, for the first 
time, a detailed hypothesis on the atomic level for the self-reproduction of 
genetic material. 

We first discuss the chemical and physical-chemical data which show that 
DNA is a long fibrous molecule. Next we explain why crystallographic evidence 
suggests that the structural unit of DNA consists not of one but of two polynucle- 
otide chains. We then discuss a stereochemical model which we believe satis- 
factorily accounts for both the chemical and crystallographic data. In conclusion 
we suggest some obvious genetical implications of the proposed structure. A pre- 
liminary account of some of these data has already appeared in Nature (Watson 
and Crick, 1953a, 1953b). 

I. Evidence for the Fibrous Nature of DNA 

The basic chemical formula of DNA is now well established. As shown in 
Figure 1 it consists of a very long chain, the backbone of which is made up of 
alternate sugar and phosphate groups, joined together in regular 3' 5' phosphate 
di-ester linkages. To each sugar is attached a nitrogenous base, only four dif- 
ferent kinds of which are commonly found in DNA. Two of these — adenine and 
guanine — are purines, and the other two — thymine and cytosine — are py- 
rimidines. A fifth base, 5-methyl cytosine, occurs in smaller amounts in certain 
organisms, and a sixth, 5-hydroxy-methyl-cytosine, is found instead of cytosine 
in the T even phages (Wyatt and Cohen, 1952). 



^Aided by a Fellowship from the National Foundation for Infantile Paralysis. 

Reprinted by permission of the authors and the Long Island 

Biological Association from Cold Spring Harbor Symposia on 

Quantitative Biology, 18, 123-131 (1953). 

193 



J. D. WATSON AND F. H. C. CRICK 



D.N. A. 



BASE SUGAR 

\ 

PHOSPHATE 

BASE SUGAR 

\ 

PHOSPHATE 

/ 

BASE SUGAR 

\ 

PHOSPHATE 

/ 

BASE SUGAR 

\ 

PHOSPHATE 

/ 

BASE SUGAR 

\ 

PHOSPHATE 



Figure 1. Chemical formula (diagrammatic) of a single chain of desox3Tibonu- 
cleic acid. 

It should be noted that the chain is unbranched, a consequence of the regular 
internucleotide linkage. On the other hand the sequence of the different nucleo- 
tides is, as far as can be ascertained, completely irregular. Thus, DNA has some 
features which are regular, and some which are irregular. 

A similar conception of the DNA molecule as a long thin fiber is obtained from 
physico-chemical analysis involving sedimentation, diffusion, light scattering, 



194 



THE STRUCTURE OF DNA 

and viscosity measurements. These techniques indicate that DNA is a very 
asymmetrical structure approximately 20 A wide and many thousands of 
angstroms long. Estimates of its molecular weight currently center between 5 X 
10" and 10^ (approximately 3 X 10^ nucleotides). Surprisingly each of these 
measurements tend to suggest that the DNA is relatively rigid, a puzzhng finding 
in view of the large number of single bonds (5 per nucleotide) in the phosphate- 
sugar backbone. Recently these indirect inferences have been confirmed by 
electron microscopy. Employing high resolution techniques both Williams (1952) 
and Kahler et al. (1953) have observed, in preparations of DNA, very long thin 
fibers with a uniform width of approximately 15-20 A. 

II. Evidence for the Existence of Two Chemical Chains 

IN THE Fiber 

This evidence comes mainly from X-ray studies. The material used is the 
sodium salt of DNA (usually from calf thymus) which has been extracted, puri- 
fied, and drawn into fibers. These fibers are highly birefringent, show marked 
ultraviolet and infrared dichroism (Wilkins et al., 1951 ; P>aser and Eraser, 1951), 
and give good X-ray fiber diagrams. From a preliminary study of these, Wilkins, 
Franklin and their co-workers at King's College, London (Wilkins et al, 1953; 
Franklin and Gosling 1953a, b and c) have been able to draw certain general con- 
clusions about the structure of DNA. Two important facts emerge from their 
work. They are : 

(1) Two distinct forms of DNA exist. Firstly a crystalline form, Structure A, 
(Figure 2) which occurs at about 75 per cent relative humidity and contains ap- 
proximately 30 per cent water. At higher humidities the fibers take up more 
water, increase in length by about 30 per cent and assume Structure B (Figure 3). 
This is a less ordered form than Structure A, and appears to be paracrystalline ; 
that is, the individual molecules are all packed parallel to one another, but are not 
otherwise regularly arranged in space. In Table 1, we have tabulated some of the 
characteristic features which distinguish the two forms. The transition from A to 
B is reversible and therefore the two structures are likely to be related in a simple 
manner. 

Table 1. 
(From Franklin and Gosling, 1953a, b and c) 











Number of 






Location of 




nucleotides 


Degree of 


Repeat distance 


first equatorial 


Water 


within unit 


orientation 


along fiber axis 


spacing 


content 


cell 


Structure A Crystalline 


28 A 


18 A 


30% 


22-24 


Structure B Paracrystalline 


34 A 


22-24 A 


>30% 


20 (?) 



(2) The crystallographic unit contains two polynucleotide chains. The argument 
is crystallographic and so will only be given in outline. Structure B has a very 
strong 3.4 A reflexion on the meridian. As first pointed out by Astbury (1947), 
this can only mean that the nucleotides in it occur in groups spaced 3.4 A apart in 



195 



J. D. WATSON AND F. H. C. CRICK 

the fiber direction. On going from Structure B to Structure A the fiber shortens 
by about 30 per cent. Thus in Structure A the groups must be about 2.5 per cent 
A apart axially. The measured density of Structure A, (Frankhn and Goshng, 




Figure 2. X-ray fiber diagram of structure A of desoxyribonucleic acid. (H. M. 
F. Wilkins and H. R. Wilson, unpub.) 

1953c) together with the cell dimensions, shows that there must be two nucleo- 
tides in each such group. Thus it is very probable that the crystallographic unit 
consists of two distinct polynucleotide chains. Final proof of this can only come 
from a complete solution of the structure. 

Structure A has a pseudo-hexagonal lattice, in which the lattice points are 22 A 
apart. This distance roughly corresponds with the diameter of fibers seen in the 
electron microscope, bearing in mind that the latter are quite dry. Thus it is 
probable that the crystallographic unit and the fiber are the one and the same. 

III. Description of the Proposed Structure 
Two conclusions might profitably be drawn from the above data. Firstly, the 



196 



THE STRUCTURE OF DMA 

structure of DNA is regular enough to form a three dimensional crystal. This 
is in spite of the fact that its component chains may have an irregular sequence of 
purine and pyrimidine nucleotides. Secondly, as the structure contains two 




Figure 3. X-ray fiber diagram of Structure B of desoxyribonucleic acid. 
Franklin and R. Gosling, 1953a.) 



(R. E. 



chains, these chains must be regularly arranged in relation to each other. 

To account for these findings, we have proposed (Watson and Crick, 1953a) a 
structure in which the two chains are coiled round a common axis and joined to- 
gether by hydrogen bonds between the nucleotide bases (see Figure 4). Both 
chains follow right handed helices, but the sequences of the atoms in the phos- 
phate-sugar backbones run in opposite directions and so are related by a dyad 
perpendicular to the helix axis. The phosphates and sugar groups are on the out- 
side of the helix whilst the bases are on the inside. The distance of a phosphorus 
atom from the fiber axis is 10 A. We have built our model to correspond to Struc- 
ture B, which the X-ray data show to have a repeat distance of 34 A in the fiber 
direction and a very strong reflexion of spacing 3.4 A on the meridian of the X-ray 



197 



J. D. WATSON AND F. H. C. CRICK 




Figure 4. This figure is diagrammatic. The two ribbons symboHze the two 
phosphate-sugar chains and the horizontal rods the pairs of bases holding the 
chain together. The vertical line marks the fiber axis, 

pattern. To fit these observations our structure has a nucleotide on each chain 
every 3.4 A in the fiber direction, and makes one complete turn after 10 such 
intervals, i.e., after 34 A. Our structure is a well-defined one and all bond dis- 
tances and angles, including van der Waal distances, are stereochemically accept- 
able. 



198 



THE STRUCTURE OF DNA 



ADENINE ° THYMINE 




4\ 



5 ii 



I I I 



Figure 5. Pairing of adenine and thymine. Hydrogen bonds are shown dotted. 
One carbon atom of each sugar is shown. 



CUANIM 



CYTOSINE 




Figure 6. Pairing of guanine and cytosine. Hydrogen bonds are shown dotted. 
One carbon atom of each sugar is shown. 

The essential element of the structure is the manner in which the two chains 
are held together by hydrogen bonds between the bases. The bases are perpendic- 
ular to the fiber axis and joined together in pairs. The pairing arrangement is 
very specific, and only certain pairs of bases will fit into the structure. The basic 
reason for this is that we have assumed that the backbone of each polynucleotide 



199 



J. D. WATSON AND F. H. C. CRICK 

chain is in the form of a regular helix. Thus, irrespective of which bases are 
present, the glucosidic bonds (which join sugar and base) are arranged in a regu- 
lar manner in space. In particular, any two glucosidic bonds (one from each 
chain) which are attached to a bonded pair of bases, must always occur at a fixed 
distance apart due to the regularity of the two backbones to which they are joined. 
The result is that one member of a pair of bases must always be a purine, and 
the other a pyrimidine, in order to bridge between the two chains. If a pair con- 
sisted of two purines, for example, there would not be room for it; if of two 
pyrimidines they would be too far apart to form hydrogen bonds. 

In theory a base can exist in a number of tautomeric forms, differing in the 
exact positions at which its hydrogen atoms are attached. However, under physi- 
ological conditions one particular form of each base is much more probable than 
any of the others. If we make the assumption that the favored forms always oc- 
cur, then the pairing requirements are even more restrictive. Adenine can only 
pair with thymine, and guanine only with cytosine (or 5-methyl-cytosine, or 5- 
hydroxy-methyl-cytosine). This pairing is shown in detail in Figures 5 and 6. If 
adenine tried to pair with cytosine it could not form hydrogen bonds, since there 
would be two hydrogens near one of the bonding positions, and none at the other, 
instead of one in each. 

A given pair can be either way round. Adenine, for example, can occur on 
either chain, but when it does its partner on the other chain must always be 
thymine. This is possible because the two glucoside bonds of a pair (see Figures 5 
and 6) are symmetrically related to each other, and thus occur in the same po- 
sitions if the pair is turned over. 

It should be emphasized that since each base can form hydrogen bonds at a 
number of points one can pair up isolated nucleotides in a large variety of ways. 
Specific pairing of bases can only be obtained by imposing some restriction, and 
in our case it is in a direct consequence of the postulated regularity of the phos- 
phate-sugar backbone. 

It should further be emphasized that whatever pair of bases occurs at one par- 
ticular point in the DNA structure, no restriction is imposed on the neighboring 
pairs, and any sequence of pairs can occur. This is because all the bases are flat, 
and since they are stacked roughly one above another like a pile of pennies, it 
makes no difference which pair is neighbor to which. 

Though any sequence of bases can fit into our structure, the necessity for spe- 
cific pairing demands a definite relationship between the sequences on the two 
chains. That is, if we knew the actual order of the bases on one chain, we could 
automatically write down the order on the other. Our structure therefore consists of 
two chains, each of which is the complement of the other. 

IV. Evidence in Favor of the Complementary Model 
The experimental evidence available to us now offers strong support to our 

model though we should emphasize that, as yet, it has not been proved correct. 

The evidence in its favor is of three types: 

(1) The general appearance of the X-ray picture strongly suggests that the 



200 



THE STRUCTURE OF DNA 

basic structure is helical (Wilkins et al., 1953; Franklin and Gosling, 1958a). If 
we postulate that a helix is present, we immediately are able to deduce from the 
X-ray pattern of Structure B (Figure 3), that its pitch is 34 A and its diameter ap- 
proximately 20A. Moreover, the pattern suggests a high concentration of atoms 
on the circumference of the helix, in accord with our model which places the 
phosphate sugar backbone on the outside. The photograph also indicates that 
the two polynucleotide chains are not spaced equally along the fiber axis, but are 
probably displaced from each other by about three-eighths of the fiber axis 
period, an inference again in qualitative agreement with our model. 

The interpretation of the X-ray pattern of Structure A (the crystalline form) is 
less obvious. This form does not give a meridional reflexion at 3.4 A, but instead 
(Figure 2) gives a series of reflexions around 25° off the meridian at spacings be- 
tween 3 A and 4 A. This suggests to us that in this form the bases are no longer 
perpendicular to the fiber axis, but are tilted about 25° from the perpendicular 
position in a way that allows the fiber to contract 30 per cent and reduces the 
longitudinal translation of each nucleotide to about 2.5 A. It should be noted 
that the X-ray pattern of Structure A is much more detailed than that of Struc- 
ture B and so if correctly interpreted, can yield more precise information about 
DNA. Any proposed model for DNA must be capable of forming either Struc- 
ture A or Structure B and so it remains imperative for our very tentative interpre- 
tation of Structure A to be confirmed. 

(2) The anomolous titration curves of undegraded DNA with acids and bases 
strongly suggests that hydrogen bond formation is a characteristic aspect of- DNA 
structure. When a solution of DNA is initially treated with acids or bases, no 
groups are titratable at first between pH 5 and pH 11.0, but outside these limits 
a rapid ionization occurs (Gulland and Jordan, 1947; Jordan, 1951). On back 
titration, however, either with acid from pH 12 or with alkali from pH 2^, a dif- 
ferent titration curve is obtained indicating that the titratable groups are more 
accessible to acids and bases than is the untreated solution. Accompanying the 
initial release of groups at pH 1 1.5 and in the range pH 3.5 to pH 4.5 is a marked 
fall in the viscosity and the disappearance of strong flow birefringence. While 
this decrease was originally thought to be caused by a reversible depolymerization 
(Vilbrandt and Tennent, 1943), it has been shown by Gulland, Jordan and Taylor 
(1947) that this is unlikely as no increase was observed in the amount of sec- 
ondary phosphoryl groups. Instead these authors suggested that some of the 
groups of the bases formed hydrogen bonds between different bases. They 
were unable to decide whether the hydrogen bonds linked bases in the same or in 
adjacent structural units. The fact that most of the ionizable groups are orig- 
inally inaccessible to acids and bases is more easily explained if the hydrogen 
bonds are between bases within the same structural unit. This point would defi- 
nitely be established if it were shown that the shape of the initial titration curve 
was the same at very low DNA concentrations, when the interaction between 
neighboring structural units is small. 

(3) The analytical data on the relative proportion of the various bases show 
that the amount of adenine is close to that of thymine, and the amount of guanine 



201 



J. D. WATSON AND F. H. C. CRICK 

close to the amount of cytosine + 5-methyl cytosine, although the ratio of ade- 
nine to guanine can vary from one source to another (Chargaff, 1951; Wyatt, 
1952). In fact as the techniques for estimation of the bases improve, the ratios of 
adenine to thymine, and guanine to cytosine + 5-methyl cytosine appear to grow 
very close to unity. This is a most striking result, especially as the sequence of 
bases on a given chain is likely to be irregular, and suggests a structure involving 
paired bases. In fact, we believe the analytical data offer the most important 
evidence so far available in support of our model, since they specifically support 
the biologically interesting feature, the presence of complementary chains. 
We thus believe that the present experimental evidence justifies the working 
hypothesis that the essential features of our model are correct and allows us to 
consider its genetic possibilities. 

V. Genetical Implications of the Complementary Model 

As a preliminary we should state that the DNA fibers from which the X-ray 
diffraction patterns were obtained are not artifacts arising in the method of pre- 
paration. In the first place, Wilkins and his co-workers (see Wilkins et al., 1953) 
have shown that X-ray patterns similar to those from the isolated fibers can be 
obtained from certain intact biological materials such as sperm head and bacterio- 
phage particles. Secondly, our postulated model is so extremely specific that we 
find it impossible to believe that it could be formed during the isolation from 
living cells. 

A genetic material must in some way fulfil two functions. It must duplicate 
itself, and it must exert a highly specific influence on the cell. Our model for 
DNA suggests a simple mechanism for the first process, but at the moment we 
cannot see how it carries out the second one. We believe, however, that its speci- 
ficity is expressed by the precise sequence of the pairs of bases. The backbone of 
our model is highly regular, and the sequence is the only feature which can carry 
the genetical information. It should not be thought that because in our structure 
the bases are on the "inside," they would be unable to come into contact with 
other molecules. Owing to the open nature of our structure they are in fact fairly 
accessible. 

A Mechanism for DNA Replication 

The complementary nature of our structure suggests how it duplicates itself. 
It is difficult to imagine how like attracts like, and it has been suggested (see 
Pauling and Delbruck, 1940; Friedrich-Freksa, 1940; and Muller, 1947) that self 
duplication may involve the union of each part with an opposite or comple- 
mentary part. In these discussions it has generally been suggested that protein 
and nucleic acid are complementary to each other and that self replication involves 
the alternate syntheses of these two components. We should like to propose 
instead that the specificity of DNA self replication is accomplished without 
recourse to specific protein synthesis and that each of our complementary DNA 
chains serves as a template or mould for the formation onto itself of a new com- 
panion chain. 



202 



THE STRUCTURE OF DNA 

For this to occur the hydrogen bonds Hnking the complementary chains must 
break and the two chains unwind and separate. It seems likely that the single 
chain (or the relevant part of it) might itself assume the helical form and serve as 
a mould onto which free nucleotides (strictly polynucleotide precursors) can 
attach themselves by forming hydrogen bonds. We propose that polymerization 
of the precursors to form a new chain only occurs if the resulting chain forms the 
proposed structure. This is plausible because steric reasons would not allow 
monomers "crystallized" onto the first chain to approach one another in such a 
way that they could be joined together in a new chain, unless they were those 
monomers which could fit into our structure. It is not obvious to us whether a 
special enzyme would be required to carry out the polymerization or whether the 
existing single helical chain could act effectively as an enzyme. 

Difficulties in the Replication Scheme 

While this scheme appears intriguing, it nevertheless raises a number of 
difficulties, none of which, however, do we regard as insuperable. The first 
difficulty is that our structure does not differentiate between cytosine and 5- 
methyl cytosine, and therefore during replication the specificity in sequence in- 
volving these bases would not be perpetuated. The amount of 5-methyl cytosine 
varies considerably from one species to another, though it is usually rather small 
or absent. The present experimental results (Wyatt, 1952) suggest that each 
species has a characteristic amount. They also show that the sum of the two 
cytosines is more nearly equal to the amount of guanine than is the amount of 
cytosine by itself. It may well be that the difference between the two cytosines is 
not functionally significant. This interpretation would be considerably strength- 
ened if it proved possible to change the amount of 5-methyl cytosine in the DNA 
of an organism without altering its genetical make-up. 

The occurrence of 5-hydroxy-methyl-cytosine in the T even phages (Wyatt 
and Cohen, 1952) presents no such difficulty, since it completely replaces cytosine, 
and its amount in the DNA is close to that of guanine. 

The second main objection to our scheme is that it completely ignores the role 
of the basic protamines and histones, proteins known to be combined with DNA 
in most living organisms. This was done for two reasons. Firstly, we can for- 
mulate a scheme of DNA reproduction involving it alone and so from the view- 
point of simplicity it seems better to believe (at least at present) that the genetic 
specificity is never passed through a protein intermediary. Secondly, we know 
almost nothing about the structural features of protamines and histones. Our 
only clue is the finding of Astbury (1947) and of Wilkins and Randall (1953) 
that the X-ray pattern of nucleoprotamine is very similar to that of DNA alone. 
This suggests that the protein component, or at least some of it, also assumes a 
helical form and in view of the very open nature of our model, we suspect that 
protein forms a third helical chain between the pair of polynucleotide chains 
(see Figure 4). As yet nothing is known about the function of the protein; 
perhaps it controls the coiling and uncoiling and perhaps it assists in holding 
the single polynucleotide chains in a helical configuration. 



203 



J. D. WATSON AND F. H. C. CRICK 

The third difficulty involves the necessity for the two complementary chains 
to unwind in order to serve as a template for a new chain. This is a very funda- 
mental difficulty when the two chains are interlaced as in our model. The two 
main ways in which a pair of helices can be coiled together have been called 
plectonemic coiling and paranemic coiling. These terms have been used by 
cytologists to describe the coiling of chromosomes (Huskins, 1941; for a review 
see Manton, 1950). The type of coiling found in our model (see Figure 4) is 
called plectonemic. Paranemic coiling is found when two separate helices are 
brought to lie side by side and then pushed together so that their axes roughly 
coincide. Though one may start with two regular helices the process of pushing 
them together necessarily distorts them. It is impossible to have paranemic 
coiling with two regular simple helices going round the same axis. This point 
can only be clearly grasped by studying models. 

There is of course no difficulty in "unwinding" a single chain of DNA coiled 
into a helix, since a polynucleotide chain has so many single bonds about which 
rotation is possible. The difficulty occurs when one has a pair of simple helices 
with a common axis. The difficulty is a topological one and cannot be sur- 
mounted by simple manipulation. Apart from breaking the chains there are 
only two sorts of ways to separate two chains coiled plectonemically. In the 
first, one takes hold of one end of one chain, and the other end of the other, and 
simply pulls in the axial direction. The two chains slip over each other, and finish 
up separate and end to end. It seems to us highly unlikely that this occurs in 
this case, and we shall not consider it further. In the second way the two chains 
must be directly untwisted. When this has been done they are separate and side 
by side. The number of turns necessary to untwist them completely is equal 
to the number of turns of one of the chains round the common axis. For our 
structure this comes to one turn every 34 A, and thus about 150 turns per million 
molecular weight of DNA, that is per 5000 A of our structure. The problem of 
uncoiling falls into two parts: 

(1) How many turns must be made, and how is tangling avoided? 

(2) What are the physical or chemical forces which produce it? 

For the moment we shall be mainly discussing the first of these. It is not easy 
to decide what is the uninterrupted length of functionally active DNA. As a 
lower limit we may take the molecular weight of the DNA after isolation, say 
fifty thousand A in length and having about 1000 turns. This is only a lower 
limit as there is evidence suggesting a breakage of the DNA fiber during the 
process of extraction. The upper limit might be the total amount of DNA in a 
virus or in the case of a higher organism, the total amount of DNA in a chromo- 
some. For T2 this upper limit is approximately 800,000 A which corresponds to 
20,000 turns, while in the higher organisms this upper limit may sometimes be 
1000 fold higher. 

The difficulty might be more simple to resolve if successive parts of a chromo- 
some coiled in opposite directions. The most obvious way would be to have 
both right and left handed DNA helices in sequence but this seems unlikely as 
we have only been able to build our model in the right handed sense. Another 



204 



THE STRUCTURE OF DNA 

possibility might be that the long strands of right handed DNA are joined together 
by compensating strands of left handed polypeptide helices. The merits of this 
proposition are difficult to assess, but the fact that the phage DNA does not seem 
to be linked to protein makes it rather unattractive. 

The untwisting process would be less complicated if replication started at the 
ends as soon as the chains began to separate. This mechanism would produce 
a new two-strand structure without requiring at any time a free single-strand 
stage. In this way the danger of tangling would be considerably decreased as 
the two-strand structure is much more rigid than a single strand and would 
resist attempts to coil around its neighbors. Once the replicating process is 
started the presence, at the growing end of the pair, of double-stranded structures 
might facilitate the breaking of hydrogen bonds in the original unduplicated 
section and allow replication to proceed in a zipper-like fashion. 

It is also possible that one chain of a pair occasionally breaks under the strain 
of twisting. The polynucleotide chain remaining intact could then release the 
accumulated twist by rotation about single bonds and following this, the broken 
ends, being still in close proximity, might rejoin. 

It is clear that, in spite of the tentative suggestions we have just made, the 
difficulty of untwisting is a formidable one, and it is therefore worthwhile re- 
examining why we postulate plectonemic coiling, and not paranemic coiling in 
which the two helical threads are not intertwined, but merely in close apposition 
to each other. Our answer is that with paranemic coiling, the specific pairing of 
bases would not allow the successive residues of each helix to be in equivalent 
orientation with regard to the helical axis. This is a possibility we strongly 
oppose as it implies that a large number of stereochemical alternatives for the 
sugar-phosphate backbone are possible, an inference at variance to our finding, 
with stereochemical models (Crick and Watson, 1953) that the position of the 
sugar-phosphate group is rather restrictive and cannot be subject to the large 
variability necessary for paranemic coiling. Moreover, such a model would not 
lead to specific pairing of the bases, since this only follows if the glucosidic links 
are arranged regularly in space. We therefore believe that if a helical structure 
is present, the relationship between the helices will be plectonemic. 

We should ask, however, whether there might not be another complementary 
structure which maintains the necessary regularity but which is not helical. 
One such structure can, in fact, be imagined. It would consist of a ribbon-like 
arrangement in which again the two chains are joined together by specific pairs 
of bases, located 3.4 A above each other, but in which the sugar-phosphate 
backbone instead of forming a helix, runs in a straight line at an angle approx- 
imately 30° off the line formed by the pair of bases. While this ribbon-like 
structure would give many of the features of the X-ray diagram of Structure B, 
we are unable to define precisely how it should pack in a macroscopic fiber, 
and why in particular it should give a strong equatorial reflexion at 20-24 A. 
We are thus not enthusiastic about this model though we should emphasize 
that it has not yet been disproved. 

Independent of the details of our model, there are two geometrical problems 



205 



J. D. WATSON AND F. H. C. CRICK 



ADENINE 



THYMINE 




ADENINE 



CYTOSINE 




Figure 7. Pairing arrangements of adenine before (above) and after (below) 
it has undergone a tautomeric shift. 

which any model for DNA must face. Both involve the necessity for some form 
of super folding process and can be illustrated with bacteriophage. Firstly, the 
total length of the DNA within T2 is about 8 X lO^A. As its DNA is thought 
(Siegal and Singer, 1953) to have the same very large M. W. as that from other 
sources, it must bend back and forth many times in order to fit into the phage 
head of diameter 800 A. Secondly, the DNA must replicate itself without getting 
tangled. Approximately 500 phage particles can be synthesized within a single 
bacterium of average dimensions 10^ X 10^ X 2 X lO"* A. The total length of 
the newly produced DNA is some 4 X 10^ A, all of which we believe was at some 
interval in contact with its parental template. Whatever the precise mechanism 
of replication we suspect the most reasonable way to avoid tangling is to have 
the DNA fold up into a compact bundle as it is formed. 

A Possible Mechanism for Natural Mutation 

In our duplication scheme, the specificity of replication is achieved by means 
of specific pairing between purine and pyrimidine bases; adenine with thymine, 



206 



THE STRUCTURE OF DNA 

and guanine with one of the cytosines. This specificity results from our assump- 
tion that each of the bases possesses one tautomeric form which is very much 
more stable than any of the other possibilities. The fact that a compound is 
tautomeric, however, means that the hydrogen atoms can occasionally change 
their locations. It seems plausible to us that a spontaneous mutation, which as 
implied earlier we imagine to be a change in the sequence of bases, is due to a 
base occurring very occasionally in one of the less likely tautomeric forms, at 
the moment when the complementary chain is being formed. For example, 
while adenine will normally pair with thymine, if there is a tautomeric shift of 
one of its hydrogen atoms it can pair with cytosine (Figure 7). The next time 
pairing occurs, the adenine (having resumed its more usual tautomeric form) 
will pair with thymine, but the cytosine will pair with guanine, and so a change 
in the sequence of bases will have occurred. It would be of interest to know the 
precise difference in free energy between the various tautomeric forms under 
physiological conditions. 

General Conclusion 

The proof or disproof of our structure will have to come from further crystal- 
lographic analysis, a task we hope will be accomplished soon. It would be 
surprising to us, however, if the idea of complementary chains turns out to be 
wrong. This feature was initially postulated by us to account for the crystal- 
lographic regularity and it seems to us unlikely that its obvious connection with 
self replication is a matter of chance. On the other hand the plectonemic coiling 
is, superficially at least, biologically unattractive and so demands precise 
crystallographic proof. In any case the evidence for both the model and the 
suggested replication scheme will be strengthened if it can be shown unambig- 
uously that the genetic specificity is carried by DNA alone, and, on the molecular 
side, how the structure could exert a specific influence on the cell. 

References 

AsTBURY, W. T., 1947, X-Ray Studies of nucleic acids in tissues. Sym. Soc. Exp. Biol. 

/: 66-76. 
Chargaff, E., 1951, Structure and function of nucleic acids as cell constituents. Fed. 

Proc. 10: 654-659. 
Crick, F. H. C, and Watson, J. D., 1953, Manuscript in preparation. 
Franklin, R. E., and Gosling, R., 1953a, Molecular configuration in sodium thymo- 

nucleate. Nature, Lond. i 7/ : 740-741. 
1953b, Fiber diagrams of sodium thymonucleate. I. The influence of water content. 

Acta Cryst., Camb. (in press). 
1953c, The structure of sodium thymonucleate fibers. II. The cylindrically symmetri- 
cal Patterson Function. Acta Cryst., Camb. (in press). 
Fraser, M. S., and Eraser, R. D. B., 1951, Evidence on the structure of desoxyribo- 

nucleic acid from measurements with polarized infra-red radiation. Nature, Lond. 

167: 760-761. 
Friedrich-Freksa, H., 1940, Bei der Chromosomen Konjugation wirksame Krafte und 

ihre Bedeutung fur die identische Verdoppling von Nucleoproteinen. Naturwissen- 

shaften 28: 376-379. 



207 



J. D. WATSON AND F. H. C. CRICK 

GuLLAND, J. M., and Jordan, D. 0., 1946, The macromolecular behavior of nucleic acids. 

Sym. Soc. Exp. Biol. / : 56-65. 
GuLLAND, J. M., Jordan, D. 0., and Taylor, H. F. W., 1947, Electrometric titration of 

the acidic and basic groups of the desoxypentose nucleic acid of calf thymus. J. 

Chem. Soc. 1131-1141. 
HusKiNS, C. L., 1941, The coiling of chromonemata. Cold Spr. Harb. Symp. Quant. Biol. 

9: 13-18. 
Jordan, D. 0., 1951, Physiochemical properties of the nucleic acids. Prog. Biophys. 2: 

51-89. 
Kahler, H., and Lloyd, B. J., 1953, The electron microscopy of sodium desoxyribo- 

nucleate. Biochim. Biophys. Acta 10: 355-359. 
Manton, I., 1950, The spiral structure of chromosomes. Biol. Rev. S5: 486-508. 
MuLLER, H. J., 1947, The Gene. Proc. Roy. Soc. Lond. Ser. B. 134: 1-37. 
Pauling, L., and Delbrijck, M., 1940, The nature of the intermolecular forces operative 

in biological processes. Science 92: 11 -I'd 
Siegal, a., and Singer, S. J., 1953, The preparation and properties of desoxypen- 

tosenucleic acid. Biochim. Biophys. Acta 10: 311-319. 
ViLBRANDT, C. F., and Tennent, H. G., 1943, The effect of pH changes upon some 

properties of sodium thymonucleate solutions. J. Amer. Chem. Soc. 63: 1806-1809. 
Watson, J. D., and Crick, F. H. C, 1953a, A structure for desoxyribose nucleic acids. 

Nature, Lond. 171: 737-738. 
1953b, Genetical implications of the structure of desoxyribose nucleic acid. Nature, 

Lond. (in press). 
WiLKiNS, M. H. F., Gosling, R. G., and Seeds, W. E., 1951, Physical studies of nucleic 

acids — nucleic acid: an extensible molecule. Nature, Lond. 167: 759-760. 
WiLKiNs, M. H. F., and Randall, J. T., 1953, Crystallinity in sperm-heads: molecular 

structure of nucleoprotein in vivo. Biochim. Biophys. Acta 10: 192 (1953). 
WiLKiNS, M. H. F., Stokes, A. R., and Wilson, H. R., 1953, Molecular structure of 

desoxypentose nucleic acids. Nature, Lond. 171: 738-740. 
Williams, R. C, 1952, Electron microscopy of sodium desoxyribonucleate by use of a new 

freeze-drying method. Biochim. Biophys. Acta 9: 237-239. 
Wyatt, G. R., 1952, Specificity in the composition of nucleic acids. In "The Chemistry 

and Physiology of the Nucleus," pp. 201-213, N. Y. Academic Press. 
Wyatt, G. R., and Cohen, S. S., 1952, A new pyrimidine base from bacteriophage nucleic 

acid. Nature, Lond. 170: 1072. 



208 



FINE STRUCTURE OF A GENETIC REGION IN BACTERIOPHAGE 

By Seymour Benzer 

BIOPHYSICAL LABORATORY, DEPARTMENTS OF BIOLOGICAL SCIENCES AND PHYSICS, 
PURDUE UNIVERSITY, LAFAYETTE, INDIANA 

Communicated by M. Delbriick, April 6, 1955 

This paper describes a functionally related region in the genetic material of a 
bacteriophage that is finely subdivisible by mutation and by genetic recombination. 
The group of mutants resembles similar cases which have been observed in many 
organisms, usually designated as "pseudo-alleles." (See reviews by Lewis^ and 

Reprinted by permission of the author and the National Academy 
of Sciences from the Proceedings of the National Academy of 
Sciences, 41 (6), 344-354, June, 1955. 

209 



Vol. 41, 1955 GENETICS: S. BENZER 345 

Pontecorvo.-) Such cases are of special interest for their bearing on the structure 
and function of genetic determinants. 

The phenomenon of genetic recombination provides a powerful tool for separating 
mutations and discerning their positions along a chromosome. When it comes to 
very closely neighboring mutations, a difficulty arises, since the closer two mutations 
lie to one another, the smaller is the probability that recombination between them 
will occur. Therefore, failure to observe recombinant types among a finite number 
of progeny ordinarily does not justify the conclusion that the two mutations are 
inseparable but can only place an upper limit on the linkage distance between 
them. A high degree of resolution requires the examination of very many prog- 
eny. This can best be achieved if there is available a selective feature for the de- 
tection of small proportions of recombinants. 

Such a feature is offered by the case of the rll mutants of T4 bacteriophage 
described in this paper. The wild-type phage produces plaques on either of two 
bacterial hosts, B or K, while a mutant of the rll group produces plaques only on 
B. Therefore, if a cross is made between two different rll mutants, any wild-type 
recombinants which arise, even in proportions as low as 10 ~*, can be detected by 
plating on K. 

This great sensitivity prompts the question of how closely the attainable resolu- 
tion approaches the molecular limits of the genetic material. From the experi- 
ments of Hershey and Chase, ^ it appears practically certain that the genetic infor- 
mation of phage is carried in its DNA. The amount of DNA in a particle of phage 
T2 has been determined by Hershey, Dixon, and Chase^ to be 4 X 10^ nucleotides. 
The amount for T4 is similar.^ If we accept the model of DNA structure proposed 
by Watson and Crick,^ consisting of two paired nucleotide chains, this corresponds 
to a total length of DNA per T4 particle of 2 X 10^ nucleotide pairs. We wish to 
translate linkage distances, as derived from genetic recombination experiments, into 
molecular units. This cannot be done very precisely at present. It is not known 
whether all the DNA in a phage particle is indispensable genetic material. Nor is 
it known whether a phage "chromosome" (i.e., the physical counterpart of a linkage 
group identified by genetic means) is composed of a single (duplex) DNA fiber 
or whether genetic recombination is equally probable in all chromosomal regions. 
For the purpose of a rough calculation, however, these notions will be assumed to 
be true. Thus we place the total linkage map of T4 in correspondence with 2 X 10* 
nucleotide pairs of DNA. The total knoA\ai length of the three linkage groups^ in 
phage T4 amounts to some 100 units (one unit = 1 per cent recombination in a 
standard cross). In addition, there is evidence^ for roughly another 100 units of 
length connecting two of the groups. Therefore, if we assume 200 recombination 
units to correspond to 2 X 10^ nucleotide pairs, the recombination per nucleotide 
pair is 10 ~^ per cent. That is to say, given two phage mutants whose mutations 
are localized in their chromosomes at sites only one nucleotide pair apart, a cross 
between these mutants should give rise to a progeny population in which one par- 
ticle in 10^ results from recombination between the mutations (provided, of course, 
that recombhiation is possible between adjacent nucleotide pairs). This compu- 
tation is an exceedingly rough one and is only intended to indicate the order of mag- 
nitude of the scale factor. Some preliminary results are here presented of a pro- 
gram designed to extend genetic studies to the molecular (nucleotide) level. 

210 



346 GENETICS: S. BENZER Proc. N. A. S. 

r Mutants. — The wild-type phages T2, T4, and T6 produce small plaques with 
rough edges when plated on strain B of Escherichia coli. From sectors of clearing 
in these plaques, mutants can be readily isolated which produce large, sharp-edged 
plaques (Hershey^). These mutants have been designated "r" for rapid lysis; 
they differ from the wild type by a failure to cause "lysis inhibition" on strain 
B (Doermann^"). The wild type has a selective advantage over r mutants when the 
two types grow together on B. The genetics of r mutants was studied by Hershey 
and Rotman,^^ who found three regions in the linkage map of T2 in which various 
mutations causing the r phenotype were located, including one large "cluster" of 
mutants which were shown to be genetically distinct from one another. The 
genetic study of T4 by Doermann and HilF showed r regions corresponding to two 
of those in T2. T6 also has at least two such r regions. 

The rll Group. — For all three phages, T2, T4, and T6, the r mutants can be sep- 
arated into groups on the basis of their behavior on strains other than B. This 
paper will be concerned only with one group, which will be called the "rll group." 
Mutants of the rll group are distinguished from those of other groups, and from 
wild type, by a failure to produce plaques on certain lysogenic strains^^ of E. coli 
which carry phage X. As shown in Table 1, a mutant of the rll group produces 

TABLE 1 

Phenotypes (Plaque Morphology) of T4 Wild and rll 
Mutant Plated on Various Hosts 



'— ■ — ■ • Host Strain . 

E. coli E. coli E. coli 

B K12S K12S (X) 

T4 wild type Wild Wild Wild 

T4 rll mutant r Type Wild 

r-type plaques on strain B, wild-type plaques on strain K12S (nonlysogenic strain 
sensitive to X), and no plaques on K12S (X) (derived from K12S by lysogenization 
with X). The wild-type phage produces similar plaques on all three strains. In 
the case of T4, with which we shall be concerned in this paper, the efficiencies of 
plating are approximately equal on the three strains, except, of course, for rll on 
K12S (X). The three bacterial strains will be here designated as "B," "S," and 
"K." 

Approximately two-thirds of the independently arising r mutants isolated on 
B are of the rll type. This group includes the "cluster" of r mutants of T2 de- 
scribed by Hershey and Rotman and the r47 and r51 mutants described by Doer- 
mann and Hill in the corresponding map region of T4 but does not include r mu- 
tants located outside that region. Similarly, all newly isolated mutants showing 
the rH character have turned out to fall within the same region, as indicated in 
Figure 1. 

The properties of the rll group are especially favorable for detailed genetic 
study. An rll mutant has three different phenotypes on the three host strains 
(Table 1): (1) altered plaque morphology on B, (2) indistinguishable from wild 
type on S, and (3) unable to produce plaques on K. These properties are all useful. 
By virtue of their altered plaque type on B, r mutants are readily isolated, and 
those of the rll group are identified by testing on K. Where it is desired to avoid 
a selective disadvantage compared with wild type, e.g., in measuring mutation 

211 



Vol. 41, 1955 GENETICS: S. BENZER 347 

rates, S can be used as a nondiscriminating host. The failure of rll mutants to 
plate on K enables one to detect very small proportions of wild-type particles due 
to reversion or due to recombination between different rll mutants. 

m r r tu 

42 47 51 41 

1 \mU 1— 



p HD% ^^ r^^\OT\ 




Fig. 1. — Partial linkage map of T4 (Doermann), indicating the location of the rll region, 
m and tu designate "minute plaque" and "turbid plaque" mutations. The circular inset shows, 
diagrammatically, the corresponding dimensions of the DNA chain magnified 1,000 diameters. 

Fate of rll Mutants in K. — Wild-type and rll mutants adsorb equally well to 
strains S and K. Whereas the wild type provokes lysis and liberation of a burst of 
progeny on both strains, the rll mutant grows normally only on S. Infection of K 
with an rll mutant provokes very little (and/or very late) lysis, although all in- 
fected cells are killed. The block in growth of rll mutant is associated with the 
presence of the carried phage X. The reason for this association is unknown. 

Quantitative Differences in Phenotype. — While all rll mutants show the same 
phenotypic effect of poor multiplication on K, they differ in the degree of this 
effect. A certain proportion of K infected with rll actually liberates some prog- 
eny, which can be detected by plating the infected cells on B. The fraction of 
infected cells yielding progeny defines a "transmission coefficient" characteristic 
of the mutant. The transmission coefficient is insensitive to the multiplicity of in- 
fection but depends strongly upon the physiological state of the bacteria (K) and 
upon temperature. Under given conditions, however, the coefficient can be 
used as a com.parative index of degree of phenotypic effect, a "leaky" mutant having 
a high coefficient. As can be seen in Table 2, a wide range of values is found. 







TABLE 2 




Properties of T4 Mutants of the 


rll Group* 


Mutant 


Map 


Transmission 


Reversion 


Number 


Position 


Coefficient 


Index 
(units of 10-«) 


r47 





0.03 


<0.01 


rl04 


1.3 


.91 


<1 


rlOl 


2.3 


.03 


4.5 


rl03 


2.9 


.02 


<0.2 


rl05 


3.4 


.02 


1.8 


rl06 


4.9 


.55 


<1 


rSl 


6.7 


.02 


170 


rl02 


8.3 


.02 


<0.01 



* Three parameters are given for each mutant. The map position is computed from the sum of the nearest 
intervals shown in Figure 2 and is given in percentage recombination units, taking the position of r47 as zero. The 
"transmission coefficient" is a measure of phenotypic effect determined by infecting bacteria K with the mutant 
in question and is given as the fraction of such infected cells yielding plaques on strain B. The "reversion index" 
is the average fraction of wild-type particles arising in lysates of the mutant grown from a small inoculum on a non- 
selective host. 

Plaques on K. — Some rll mutants produce no plaques on K, even when as many 
as 10* particles (as measured by plaque count on B) of a stock are plated. Other 
rll mutants, however, produce various proportions of plaques on K. When the 

212 



348 GENETICS: S. BENZER Proc. N. A. S. 

plaques appearing on K are picked and retested, they fall into three categories: 
(1) a type which, like the original mutant, produces very few plaques on K and r- 
type plaques on B; (2) a type which produces plaques (often smaller than wild 
type) on K with good efficiency but r-type plaques on B; and (3) a type indis- 
tinguishable from the original wild. These three types are understood to be due 
to the following: (1) "leaking" effects, i.e., ability of the mutant to grow slightly 
on K, so that there is a chance for a few visible plaques to form; (2) a mutation 
which partially undoes the effect of the rll mutation, so that multiplication in K 
is possible, but the full wild phenotype is not achieved; and (3) apparent reverse 
mutation, which may or may not be genuine, to the original wild type. 

The proportion of each type occurring in a stock is characteristic and reproducible 
for a particular rll mutant but differs enormously from one rll mutant to another. 
There is no evident correlation in the rates of occurrence of the three types. 

Reversion Rates of rll Mutants. — Reversion of r mutants to a form indistinguish- 
able from wild type was demonstrated by Hershey,^ who made use of the selective 
advantage of wild type on B to enrich its proportion in serial transfers. Given the 
inability of rll mutants to produce plaques on K, such reversions are easily de- 
tected, even in very small proportion. An index to the frequency of reversion of a 
particular rll mutant can be obtained by preparing a lysate from a small inoculum 
(about 100 particles, say, so that there is very little chance of introducing a wild- 
type particle present in the stock). If S is used as the host, both rll mutant and 
any reversions which arise can multiply with little selection, as shown by control 
mixtures. The average fraction of wild-type particles present in several lysates is 
an index which can be shown to be roughly proportional to the probability of re- 
version per duplication of the rll mutant. Under the conditions of measurement 
the index is of the order of 10-20 times the probability of reversion per duplication. 
The plaques appearing on K must be tested by picking and replating on B. This 
eliminates the "spurious" plaques produced by partial reversions and by leaky 
mutants, which show up as r type on B. As may be seen in Table 2, the reversion 
indices for rll mutants vary over a very wide range. One mutant has been found 
which reverts 10 times more frequently than r51, so that the reversion rates cover 
a known range of over 10^-fold. 

It has not been proved that these apparent reversions constitute a genuine re- 
turn to the original wild type. However, the possibility of suppressor mutations 
distant from the site of the rll mutation has been ruled out by backcrosses to the 
original wild type. Krieg'' found very few, if any, r-type recombinants in back- 
crosses of several reversions, localizing the reverse changes to within a few tenths 
of a per cent linkage distance from the original rll mutations. One case of "par- 
tial reversion" has also been tested by backcrossing, and failure to observe rll- 
type recombinants localized the "partial reverse mutation" to within the rll region. 

Mapping of the rll Region. — A cross between two rll mutants is made by infect- 
ing a culture of B with equal multiplicities (three per bacterium) of each type. 
The yield after lysis contains the two parental types and, if the parents are geneti- 
cally distinct, two recombinant types, the double mutant and wild type. In the 
average yield from many cells, the recombinant types occur in equal numbers.'' 
In all cases thus far tested, double rll mutants, like single mutants, do not produce 
plaques on K. On the assumption that this is generally true, the proportion of 

213 



Vol. 41, 1955 



GENETICS: S. BENZER 



349 



^'oe' 



S^ 



-J21 



recombinants in the yield can be measured simply by doubling the ratio of the 
plaque count on K (which registers only the wild recombinant) to the count on B 
(which registers all types). The percentage of wild type thus measured agrees 
well with a direct count of plaque types on B. 

In this way, a series of six rll mutants of T4 (the first six isolated — not selected 
in any way) have been crossed with each other and with r47 and rol (kindly supplied 
by A. H. Doermann) in 23 of the 28 possible pairs. The results of these crosses are 
given in Figure 2 and are compatible with the indicated seriation of the mutants. 
The distances are only roughly additive; there is some systematic deviation in the 
sense that a long distance tends to be smaller than the sum of its component shorter 
ones. Part of this discrepancy is accounted for by the Visconti-Delbruck correc- 
tion for multiple rounds of mating.'^ Reversion rates were small enough to be 
negligible in these crosses. Thus, while all rll mutants in this set fall into a small 
portion of the phage linkage map, it is possible to seriate them unambiguously, 
and their positions within the region are 
well scattered. 

Tests for Pseudo-allelism. — The func- 
tional relatedness of two closely linked 
mutations causing similar defects may 
be tested by constructing diploid heter- 
ozygotes containing the two mutations 
in different configurations.'- ^ The as 
form, with both mutations in one 
chromosome, usually behaves as wild 
type, since the second chromosome sup- 
plies an intact functional unit (or units). 
However, the trans form, containhig 
one of the mutations in each chromo- 
some, may or may not produce the 
wild phenotype. If it does, it is 
concluded that the two mutations in 
question are located in separate functional units. 

In applying this test to the rll mutants, the diploid heterozygote can be simulated 
l)y a mixed infection with two kinds of phage. The rll phenotype is a failure to 
lyse K, whereas the wild phenotype is to cause lysis. If K is mixedly infected with 
wild type and rll mutant, the cells lyse, liberating both types of phage. Thus 
the presence of wild type in the cell supplies the function which is defective in rll 
type, and the rll mutation can be considered "recessive." Although it has not yet 
been tested, the cis configuration of double rll mutant plus wild type is also pre- 
sumed to produce lysis in all cases. The trans configuration is obtained by infecting 
K with the pair of rll mutants in question. This is found to give lysis or not, de- 
depending upon which rll mutants compose the pair. The results are summarized 
by the dotted line in Figure 2, indicating a division of the rll region into two seg- 
ments. If both mutants belong to the same segment, mixed infection of K gives 
the mutant phenotype (very few cells lyse). If the two mutants belong to different 
segments, extensive lysis occurs with liberation of both infecting types (and recom- 
binants). These results are summarized in Figure 3. Thus, on the basis of this 



Fig. 2.- — Larger-scale map of eight rll mu- 
tants, including Doermann's r47 and r51. 
Newly isolated mutants are numbered starting 
with /W. The recombination value (in per 
cent) for each cross is obtained by plating the 
progeny on K and on B and doubling the ratio 
of plaque count on K to count on B. 



214 



350 



GENETICS: S. BENZER 



Proc. N. a. S. 



test, the two segments of the rll region correspond to independent functional 
units. 

Actually, for mixed infection of K with two (nonleaky) mutants of the same seg- 
ment, a very small proportion of the cells do lyse and liberate wild recombinants, 
that proportion increasing with the linkage distance between the mutations. For 
two rll mutants separated by 1 per cent linkage distance (measured by a standard 
cross on B) the proportion of mixedly infected K yielding any wild particles is 
about 0.2 per cent. 

This value has bearing upon the effect upon K/B values of the heterozygous phage par- 
ticles which arise in a cross between two rll mutants on B. In such a cross between closely 
linked rll mutants, the progeny should include about 2 per cent of particles containing a 
trans configuration heterozygous piece. ^^ When one of these is plated on K, there is a cer- 
tain chance that a wild recombinant may form in the first cycle of infection, leading to pro- 
duction of a plaque. If it is assumed that these are no more likely to do so than a mixed 
infection of K with two complete mutant particles, it can be concluded that the effect of 
these heterozygous particles upon the count on K is negligible, provided that both rll mu- 
tants belong to the same segment. For mutants in different segments, however, the "effi- 
ciency" of the heterozygous particles should 
be much greater, and recombination values 
measured by the K/B method should run 
considerably higher than the true values. 
The recombination values in Figure 2 for 
crosses which transgress the segmental 
divide are probably subject to some correc- 
tion for this reason. 



active 



-i— f- 



active 



» i I 



presumed 
octive 



4+ 



-N- 



-4— t- 



-f-H 



inactive 



inactive 



active 



Fig. 3. — Summarj^ of tests for "position-effect 
pseudo-allelism" of rll mutants. Each diagram 
represents a diploid heterozygote as simulated by 
mixed infection of a bacterium (K) with two 
types of phage containing the indicated muta- 
tions. Active means extensive lysis of the mix- 
edly infected cells,- inactive means very little 
lysis. The dotted line represents a dividing 
point in the rll region, the position of which is 
defined by these results. 



Rough Mapping hy Spot Test. — If a 
stock of either of two rll mutants is 
plated on K, no plaques arise; but if 
both are plated together, some bacteria 
become infected by both mutants and, 
if this leads to the occurrence of wild- 
type recombinants, plaques are pro- 
duced. If the two mutants are such 
that wild recombinants cannot arise between them (e.g., if they contain identical 
mutations), no plaques appear. A given rll mutant may thus be tested against 
several others on a single plate by first seeding the plate with K plus the mutant 
in question (in the usual soft agar top layer) and then spotting with drops contain- 
ing the other rll mutants. 

Inspection of such a plate immediately places the unknown mutant in the proper 
segment, since spotting any mutant of segment A against any mutant of segment 
B gives a very clear spot, due to the extensive lysis of mixedly infected bacteria. 
However, for a pair of mutations belonging to the same segment, plaques are pro- 
duced only by the relatively few mixedly infected bacteria which give rise to wild 
recombinants. The greater the linkage distance between the mutations, the larger 
the number of plaques that appear in the spot. A group of mutants of the same 
segment may thus be seriated by seeding one plate with each and spotting with all 
the others. Given a previously seriated group, a new mutant can thus be quickly 



215 



Vol. 41, 1955 GENETICS: S. BENZER 351 

located within the group. This method works best for mutants which are stable 
(i.e., low reversion rate) and nonleaky, so that large numbers of phage particles can 
be plated. Reversions or pronounced leaking effects obviously cause an obscuring 
background. 

This test has been applied to a large group of stable, nonleaky rll mutants. 
Their approximate locations as deduced from these tests are shown in Figure 4. 
Some of the mutants showed anomalies which made it impossible to locate them 
as members of a series. They gave very little recombination with any of the mu- 
tants located within a certain span, while behaving normally with respect to mu- 
tants located outside that span. They are indicated in Figure 4 by horizontal 
lines extending over the span. 



segment A 



I segment B v 



r r r r r r I r r 

47 104 101 103 105 106 I 51 102 

H |-HHtH lll|l I m il l I I I I I I IjiM iii ii i liii I 

I 



a b c d 

Fig. 4.— Preliminary locations of various rll mutants, based upon spot tests 

Spot tests on numerous other mutants have shown that mutants of varied re- 
version rates, transmission coefficients, and rates of "partial reversion" occur at 
scattered positions in both segments. 

Mapping of "MicroclustersJ'— The spot test enables us to pick out "mieroclus- 
ters," i.e., groups of very closely neighboring mutations. Four such groups selected 
for further study are indicated in Figure 4, and the results of mapping them are 
given in Figure 5. While some intervals show reasonably good additivity proper- 
ties, there are some mutants which give violently anomalous results. Thus in 
mic'rocluster a, r47 gives no wild recombinants (i.e., less than 1 in 10«) with any of 
the other three mutants, but two pairs of the three do show recombination. These 
results can be understood if it is assumed that each mutation extends over a certain 
length of the chromosome, and production of wild type requires recombination 
within the space between those lengths. According to this interpretation, the 
mutations would cover the lengths indicated by the bars in Figure 5. These anom- 
alies resemble those observed in the spot tests, only they are more limited in 

span. . J J 

This observation raises the question of whether there exist true "pomt muta- 
tions (i.e., involving an alteration of only one nucleotide pair) or whether all muta- 
tions involve more or less long pieces of the chromosome. It must be remembered 
that the mutants used in these experiments were selected for extreme stability 
against reversion. This procedure would be expected to enrich the proportion of 
mutants containing gross chromosomal alterations. So far as is known, the anom- 
alous cases observed could equally well be imagined to be due to double (i.e., 
two near-by "point") mutations, inversions, or deletions of the wild-type chromo- 
some. In continuing these experiments, it would seem well advised to employ 
only mutants for which some reversion is observed. 

216 



312 1 

i9i] I6e 



0.0000 
0.0000 
0.0000 

0.0000 



[0.013 
0.0371 





EOL 155 274 
1 1 1 




271 


27S 


240 




0000 


















0.035 ' 






0.01^ 


' 0.12 \ 




B 




0.017 




O.IS 






0.12 










0.10 












0.12 










0.12 












0.10 


* 












0.15 




* 






O.IS 










" 


0.15 








0.0000 
0.0000 
0.0000 
0.0000 
0.0000 
0.0000 






0.17 












* 






* 




* 






















* 





























'% 


151 




!SC 


217 PS 
/ 









0.0000 
0.0000 
0.0000 
0.0000 
0.0000 
0.0000 
















, 


0.10 


' 










0.10 














0.12 












-• 




0.15 














0.0000 
0.0000 
0.0000 
0.0000 
0.0000 




t» 






* 




* 






* 








* 





























" 


















102 1 






237 




292 




2.7 




^332 






„ 










0.0000 
0.0000 
0.0000 


' 


0.12 








o.ir 






0.0000 






' 








' 


0.23 










0.23 












0.26 






0.0000 
0.0000 
0.0000 
0.0000 
0.0000 




0.23 








* 


■ 




" 













































Fig. 5.^ — ^Maps of microclusters 
217 



Vol. 41, 1955 GENETICS: S. BENZER 353 

Discussion. — ^The set of rll mutants defines a bounded region of a linkage group 
in which mutations may occur at various locations, all the mutations leading to 
qualitatively similar phenotypic effects. The rll region would seem, therefore, to 
be functionally connected, so that mutations arising anywhere within the region 
affect the same phenotype. This effect is expressed, in case strain B is the host, 
by failure to produce lysis inhibition; in case S is the host, by no consequence; and 
in case K is the host, by inability to multiply normally. The failure of an rll mu- 
tant to mature in K can be overcome by the presence of a wild-type phage in the 
same cell. This could be understood if the function of the region in the wild-type 
"chromosome" were to control the production of a substance or substances needed 
for reproduction of this phage in K cells. 

The phenotypic test for "pseudo-allelism" leads to the division of the region into 
two functionally distinguishable segments. These could be imagined to affect two 
necessary sequential events or could go to make up a single substance the two parts 
of which must be unblemished in order for the substance to be fully active. For 
example, each segment might control the production of a specific polypeptide 
chain, the two chains later being combined to form an enzyme. While it is not 
known whether this sort of picture is applicable, a model of this kind is capable of 
describing the observed properties of the rll mutants. The map position of a mu- 
tation would localize a change in the region (and also in the "enzyme" molecule), 
the reversion rate would characterize the type of change involved in the genetic 
material, and the degree of phenotypic effect would be an expression of the degree 
of resultant change in the activity of the enzyme. A "leaky" mutant would be 
one where this latter effect was small. While no obvious correlation has yet been 
observed among these three parameters of rll mutants, one may well show up 
upon more exhaustive study. 

"Clustering" of similar mutants separable by crossing-over has been observed for 
several characters in phage by Doermann and Hill and appears to represent the rule. 
This may well be the rule in all organisms, simply because functional genetic units 
are composed of smaller recombinational and mutational elements. One would 
expect to see this effect more readily in phage because the probability of recom- 
bination per unit of hereditary material is much greater than for higher organisms. 

By extension of these experiments to still more closely linked mutations, one may 
hope to characterize, in molecular terms, the sizes of the ultimate units of genetic 
recombination, mutation, and "function." Our preliminary results suggest that 
the chromosomal elements separable by recombination are not larger than the 
order of a dozen nucleotide pairs (as calculated from the smallest non-zero recom- 
bination value) and that mutations involve variable lengths which may extend 
over hundreds of nucleotide pairs. 

In order to characterize a unit of "function," it is necessary to define what func- 
tion is meant. The entire rll region is unitary in the sense that mutations any- 
where within the region cause the rll phenotype. On the basis of phenotype tests 
of irans configuration heterozygotes, this region can be subdivided into two func- 
tionally separable segments, each of which is estimated to contain of the order of 
4 X 10^ nucleotide pairs. If one assumes that each segment has the "function" of 
specifying the sequence of amino acids in a polypeptide chain, then the specification 
of each individual amino acid can as well be considered a unitary function. It 

218 



354 GENETICS: S. BENZER Proc. N. A. S. 

would seem feasible, with this system, to extend genetic studies even to the level 
of the latter functional elements. 

Summary. — It has been discovered that the mutations in the rll region of phage 
T4 have a characteristic in common which sets them apart from the mutations 
in all other parts of the map. This characteristic is a host-range reduction, namely, 
a failure to produce plaques on a host (K) lysogenic for phage X. The mutant phage 
particles adsorb to and kill K, but normal lysis and phage release do not occur. 

All mutants with this property are located within a sharply defined portion of 
the phage linkage map. Within that region, however, their locations are widely 
scattered. An unambiguous seriation of the mutants, with roughly additive dis- 
tances, can be accomplished, except for certain anomalous cases. 

The simultaneous presence of a wild-type phage particle in K enables the multi- 
plication of rll mutants to proceed, apparently by supplying a function in which 
the mutant is deficient. A heterozygous diploid in the trans configuration is 
simulated by a mixed infection of K with two mutant types. The application of 
the phenotype test to pairs of rll mutants leads to the division of the region into 
two functionally separable segments. 

Spontaneous reversion to wild-type had been observed for most of these mutants. 
It remains to be seen whether these are genuine reversions. Each mutant reverts 
at a characteristic rate, but the rates for different mutants differ enormously. 
Partial reversions to intermediate types are also observed. 

The mutants differ greatly in degree of residual ability to grow on K. There is 
no evident correlation between map position, reversion rate, and degree of residual 
activity of the various mutants. 

The selective feature of K for wild-type recombinants offers the possibility of 
extending the recombination studies to an analysis of the fine details of the region. 

Preliminary studies of this type indicate that the units of recombination are not 
larger than the order of one dozen nucleotide pairs and that mutations may involve 
various lengths of "chromosome." 

I am much indebted to A. D. Hershey and A. H. Doermann for stocks of their 
genetically mapped mutants, to Sydney Brenner and David Krieg for stimulating 
discussion, and to Max Delbriick for his invaluable moderating influence. 

* Supported by a grant-in-aid from the American Cancer Society upon recommendation of the 
Committee on Growth of the National Research Council. 

1 E. B. Lewis, Cold Spring Harbor Symposia Quant. Biol., 16, 159-174, 1951. 

2 G. Pontecorvo, Advances in Enzymol., 13, 121-149, 1952. 

3 A. D. Hershey and M. Chase, /. Gen. Physiol, 36, 39-56, 1952. 

* A. D. Hershey, J. Di.xon, and M. Chase, /. Gen. Physiol, 36, 777-789, 1953. 
^ E. K. Volkin, personal communication. 

6 J. D. Watson and F. H. C. Crick, Cold Spring Harbor Symposia Quant. Biol, 18, 123-131, 1953. 

7 A. H. Doermann and M. B. Hill, Genetics, 38, 79-90, 1953. 

* G. Streisinger and V. Bruce, personal communication. 
9 A. D. Hershey, Genetics, 31, 620-640, 1946. 

10 A. H. Doermann, /. Bacterial, 55, 257-276, 1948. 

II A. D. Hershey and R. Rotman, Genetics, 34, 44-71, 1949. 

12 E. M. Lederberg and J. Lederberg, Genetics, 38, 51-64, 1953. 
" D. Krieg, personal communication. 
1^ N. Visconti and M. Delbruck, Genetics, 38, 5-33, 1953. 

1* A. D. Hershey and M. Chase, Cold Spring Harbor Syynposia Quant. Biol, 16, 471-479, 1951; 
C. Levinthal, Genetics, 39, 169-184, 1954. 

219 



INDUCTION OF SPECIFIC MUTATIONS WITH 5-BROMOURACIL* 

By Seymour BENZERf and Ernst FreeseJ 

BIOPHYSICAL LABORATORY, PURDUE UNIVERSITY, LAFAYETTE, INDIANA 

Communicated by M. Delbruck, December 6, 1957 

Introduction 

The hereditary characteristics of an organism occasionally undergo abrupt 
changes (mutations), and genetic techniques have traced these to alterations at 
definite locations in the genetic structure. Recently, the fineness of this genetic 
mapping has been extended to the level where the finite molecular units (nucleotides) 
of the hereditary material limit further subdivision. At this level, local details of 
the hereditary material should exert their influence; the frequency of mutation at a 
particular point should depend upon the local molecular configuration. It is there- 
fore feasible to try to correlate genetic observations with precise molecular models, 
such as the one proposed by Watson and Crick' for the structure of DNA. 

In a fine-structure study of spontaneous mutations in phage T4, the mutability 
at different points in the genetic structure was, in fact, found to be strikingly 
varied.^ To relate mutability to actual chemical structure, it would seem promising 
to employ mutagenic agents of specific types, to act selectively on particular con- 
figurations. Since the initial discovery by Muller* and Stadler^ on induct on of 
mutations with X-rays and the discovery of chemical mutagenesis by Auerbach 
and Robson^ and by Oehlkers,^ many physical agents and chemical substances have 
been found to be mutagenic in many organisms. Some mutagens act selectively; 
in particular the induced reversion from biochemically dependent to independent 
strains has been shown to depend upon the mutant and the mutagen used. (For 
chemical mutagens in bacteria see Demerec.^) A recent comprehensive review of 
this subject has been published by Westergaard.* Mutagens in some cases pro- 
duce gross chromosomal aberrations; in others the alterations are so small as to 

Reprinted by permission of the authors and the National Academy 

of Sciences from the Proceedings of the National Academy of 

Sciences, 44 (2), 112-119, February, 1958. 

220 



Vol. 44, 1958 BIOCHEMISTRY: BENZER AND FREESE 113 

be beyond the limited resolving power of genetic techniques for the organism used. 

The absence of this limitation makes phage a suitable organism for our purposes. 
There have been reports of induction of mutations in phage by ultraviolet light,* '^^ 
nitrogen mustard, '' streptomycin, '^ and proflavine. ^^ A very provocative discovery 
is that analogues of the normal bases may be built into DNA in place of the usual 
ones and also raise the mutation rate. In particular, one such analogue, 5-bromoura- 
cil, has been proven by Dunn and Smith'* to be incorporated into the DNA of 
phage (in place of thymine), and Litman and Pardee'^ have shown that it greatly 
increases the frequency with which phage mutants of various types arise. 

In the present work, this possibility of directly affecting the DNA structure is 
combined with a genetic analysis of high resolving power, to make a fine-structure 
study of mutagenesis. Our attention will be restricted to the rll region of the 
genome of phage T4. The mutational alterations arising by 5-bromouracil induc- 
tion are compared to, and shown to differ from, those which occur spontaneously. 

METHODS AND MATERIALS 

Strains: phage T4B; bacterium B {E. colt B) for the isolation of mutants and as 
plating bacterium for the determination of phage titers; S (E. coli K12S) for the 
preparation of phage stocks; K (E. coli K12S lysogenic for prophage lambda) as 
the selective strain for genetic tests. 

Media: broth 1 per cent bacto-tryptone (Difco) plus 0.5 per cent NaCl; glucose- 
salts medium;'^ sulfanilamide medium same as used by Litman and Pardee,'* except 
for higher sulfanilamide concentration (2 mg/ml) and addition of 1 /xg/ml calcium 
pantothenate, 1 Mg/ml pyridoxine, 1 Mg/ml thiamine, 1 Mg/ml uracil, and 20 Mg/ml l- 
tryptophane (tryptophane required for adsorption of phage T4B to B in synthetic 
medium). Plates contain broth plus 1.3 per cent agar (Difco) with a top layer of 
broth plus 0.7 per cent agar. 

A sample of 5-bromouracil purified by ion-exchange column, was kindly supplied 
by Dr. Rose Litman. 

Isolation of the Mutants. — Spontaneous mutants : Details on the isolation of spon- 
taneously arising r mutants of phage T4, their properties, and the methods used in 
mapping them genetically are given in earlier publications.^' " In brief, a stock of 
standard ("wild") type phage T4 (derived from a single T4 particle) is plated on B. 
Each phage particle produces a plaque containing around 10^ progeny. The 
progeny include occasional r mxUtants, which can be found by picking the plaque 
and replating its contents. In order to assure that each mutant arises by an inde- 
pendent mutational event, no more than one r mutant is isolated from any one 
plaque of the standard type. Each r mutant is replated (to free it from any contam- 
inating particles of standard type), an isolated r-type plaque is picked, and a stock 
of the mutant grown on bacteriums in broth. 

Induced mutants: These were isolated from the yield of bacteria infected and 
allowed to burst in the presence of sulfanilamide and 5-bromouracil. A culture of B 
was prepared by inoculation of 0.4 ml. of overnight culture (grown in glucose-salts 
medium) into 20 ml. of sulfanilamide medium and aeration for 3.5 hours to reach a 
cell concentration of 7 X 10* per ml. At this time, 1 mg. of 5-bromouracil and 4 X 
10* particles of T4 standard-type phage'* were introduced simultaneously. Drops 

221 



114 BIOCHEMISTRY: BENZER AND FREESE Proc. N. A. S. 

of the mixture were rapidly distributed — one drop into each of 200 tubes. After 
incubation at 37° C. (for 30 minutes) to allow the infected cells to burst, the content 
of each tube was plated on B. As is typical in this sulfanilamide medium, the 
average yield per cell was very small, of the order of one viable progeny particle per 
infected cell. Each plate contained (after incubation) from 30 to 100 plaques, in 
most cases including one or more r-type plaques. The over-all proportion of r-type 
particles among the progeny was about 2 per cent. To assure the independent 
origin of each mutant, no more than one r plaque was picked from a plate. Each 
such mutant was purified by replating, and a stock prepared on S in broth. 

Genetic Mapping of the Mutow^s.— Different r mutants of T4, although producing 
similar plaques on B, fall into groups distinguished by their behavior on a second 
host, K. Those mutants with which we are here concerned, of the rll group, do not 
produce plaques on K. This property is the key to the high resolution with which 
they can be mapped genetically. When two rll mutants are crossed, the appearance 
of any standard-type recombinants among the progeny is readily detected by plating 
on K. If standard-type progeny are produced in a cross (above the background 
rate due to spontaneous reversion of the mutants), it is concluded that the two 
mutants contain alterations at different locations in their genetic structures. 

Our objective is to compare these locations for spontaneously arising and for 5- 
bromouracil-induced mutants. The task of crossing a large number of mutants, 
two by two, to see which pairs yield recombinants is enormous. However, this 
process can be shortened by making use of a set of mutants having large alterations, 
as shown in Figure 1. Each new isolated rll mutant is crossed with each mutant of 
this set (by means of simple spot tests). By noting with which mutants of the set 
it does or does not produce standard-type recombinants, the mutation can be as- 
signed to a particular segment of the map. Thereafter, only mutants belonging to 
the same segment need be crossed with each other. Thus the number of crosses 
reciuired for analyzing a batch of mutants is greatly reduced. 

The genetic procedure has therefore been to (1) isolate many independently 
arising r-type mutants; (2) choose those of the rll group; (3) test each rll mutant 
against the mutants of Figure 1, thereby locating its mutational alteration in a 
particular segment of the map; and (4) cross the mutants belonging to the same 
segment with each other to determine which mutations share common locations. 
For the present purposes, no attempt was made to determine the order of these loca- 
tions within a map segment. 

Reversion Rates of Mutants.— ^The different rll are characterized not only by the 
positions of their mutational alterations in the map but also by differences in their 
fre(}uency of reversion to particles that resemble the standard type in their plating 
behavior on K and B. These revertants arise spontaneously during the growth of a 
given r-type phage, and their typical frequency, in a stock grown up from an 
inoculum of 100 r phages, is called the "reversion index." 

RESULTS 

Over-all Mutation Frequencies.— The ratio of the induced to the spontaneous rate 
of mutation cannot be given accurately, since different procedures of isolation were 
used. A rough estimate may be made as follows: In a broth lysate of T4B grown 
(on S) from an inoculum of a few particles up to a population of lO^", the proportion 

222 



Vol. 44, 1958 BIOCHEMISTRY: BENZER AND F REESE 115 

of spontaneously appearing r particles is typically 2 X 10"''. Under the conditions 
used for induction, on the other hand, r mutants appeared in a proportion of around 
2 per cent in the progeny from a single growth cycle. To obtain the probability of 
mutation per phage replication, these values must in each case be adjusted to ac- 
count for the accumulation of mutants during the growth of the population. This 
effect, which is roughly proportional to the logarithm of the ratio of final to initial 
phage titer, should be several fold (perhaps four times) larger for a lysate than for 
one cycle of growth. Therefore, the rate of induced over-all r mutations can be 
estimated as several hundred times the spontaneous rate. 

rl region 



cistron A 



■^ 



r H88 



r 164 



cistron B 



r 638 



r A 105 



r 196 r 187 



segment I I | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | II | 

Fig. 1. — Genetic map of the rll region of phage T4, showing the division of the map into seg- 
ments by a set of rll mutants having large alterations. Each of the two "cistrons' ' is a functional 
unit defined by the cis-trans test. 

The relative positions of the alterations have been established by their overlapping relationships 
with a much larger set of mutants. To locate the defect of a new mutant, it is crossed with each 
member of the set, and standard-type recombinants are tested for. From the results of these 
crosses the defect can be assigned to one of the segments. For example, if a mutant gives stand- 
ard-type recombinants with all except r H88, r A105, and r 164, its defect must be located in seg- 
ment 4. One which gives recombinants with all must have its defect in segment 1. Segments 
7, 9, and 1 1 are not distinguished by this set of mutants. 

Genetic Analysis of the Mutants. — Spontaneous mutants: A series of 238 spon- 
taneously arising mutants was isolated, numbered r 101 through r 338, and analyzed 
as described above. Of the mutants, 132 belonged to the rll group. Of these, 6 
were not readily mappable because of high reversion rates or leakiness (ability to 
grow feebly on K). The results of mapping the remaining 126 mutants are shown 
in Figure 2. 

The relative positions of the sites indicated within a map segment have no 
significance. Any two mutants that yield standard type progeny when crossed are 
sketched as horizontally separated, while those groups for which recombination 
could not be detected (at a level of about 0.01 per cent recombinant progeny or 
more) are assigned to the same or adjacent horizontal sites. Some mutants fail to 
give recombinants with (any of two or more) other mutants known to give recom- 
binants with each other; the alterations of such mutants are represented as hori- 
zontal bars. A striking feature of the map is the existence of certain "hot spots," 
where mutations recur with high probability. 

Induced mutants: A series of 170 r mutants, arising under the action of 5- 
bromouracil in sulfanilamide medium, was numbered N 1 through N 170. Of these, 
89 were of the rll type. Twelve of the latter were not readily mappable because 

223 



116 



BIOCHEMISTRY: BENZER AND F REESE 



Proc. N. a. S. 



SPONTANEOUS a 



INDUCED BY B a 

BROMOURACIL y 




••gmtnl (T) 



(D (D ® 



S PONTANEOUS 



PI w H H_B H SI n ca 



INDUCED BY 1 

BROMOURACIL \ / \__ 



J \ / \ / V 



® 



@ ©.©.onddj) 



CISTRON 8 

B 



R 


eversion 


Index 


( 


units of 


10-^) 


□ 


100 - 


10,000 





1 - 


100 


S 


0.005 


— 1 


■ 


no reversion detected 



c 

Fig. 2. — Genetic maps showing the locations of spontaneous and 5-bromouracil-induced 
mutational alterations in the rll region of phage T4 divided into cistron A and B. Each mutant 
is represented by a box, placed in the proper segment of the map and shaded to indicate its re- 
version index. The relative order (but not the lengths) of segments 1, 2, 3, 4, 5, 6, 8, and 10 have 
been established. Segments 7, 9, and 11, are lumped together and considered as one segment. 
The order and position of different mutations within one segment have not been determined. 

they were leaky. The results of mapping the remaining 67 mutants are shown in 
Figure 2. This set of mutants reveals "5-bromouracil hot spots" that are located at 
different positions from the spontaneous ones. 

It must be emphasized that the vertical scales in Figure 2 are very different for 
the two sets of mutants. The over-all induced mutation rate was several hundred- 



224 



Vol. 44, 1958 BIOCHEMISTRY: BENZER AND F REESE 117 

fold higher, and only half as many induced mutants are mapped; therefore, the 
occurrence of an induced mutant once represented in the induced set corresponds to 
roughly 10^ times higher mutability than the occurrence of a spontaneous mutant 
once represented in the spontaneous set. Hence, for those "hot spots" in the induced 
set for which no spontaneous mutation has been observed, the mutability with 
bromouracil is at least 10^ times larger than is the spontaneous mutability. 

Reverse Mutations. — Among the spontaneous rll mutants, not only are forward 
mutation frequencies different for different sites, but the reversion indexes as well 
cover an enormous range. Among the 132 spontaneous rll mutants, reversion in- 
dexes range as high as 5 X lO^^ down to less than IQ-^ For 19 of them, reversion 
has not been detected. 

The induced mutants, on the other hand, are more homogeneous with respect to 
reversion index. There is only one mutant of very high reversion index (N 76) and 
two non-reverting mutants (N 101, N 32). Most of the remainder have reversion 
indexes of the order of 10 ~l Notably rare are mutants containing larger aberra- 
tions (i.e., ones that fail to give standard-type recombinants with two or more 
other mutants that do themselves show recombination). 

Some preliminary experiments have been performed on the induction of reversion 
of various mutants with 5-bromouracil. For some mutants, in particular, chosen 
from some of the "5-bromouracil hot spots," increases over the spontaneous rate 
of reversion by factors as high as 10* have been observed. Some spontaneous 
mutants show no or only much smaller effects. These experiments are still in 
progress. 

DISCUSSION 

Clearly, the mutagenic effect of 5-bromouracil is not merely a general enhance- 
ment of spontaneously occurring mutations. Rather, it is a specific effect. In de- 
noting the action of a mutagen as specific, one might be comparing the effects on 
different phenotypic characteristics in the whole organism, different cistrons, or 
different locations within a single cistron. Were it not for the high resolution of our 
genetic techniques, it might have been erroneously concluded that 5-bromouracil 
acts aspecifically, since, among all the induced mutants of the r phenotype, the 
proportion of rll mutants and even the ratio of A cistron to B cistron mutants are 
quite comparable to the spontaneous values. It is only at finer resolution that the 
different effects are revealed, and it is seen that mutations are, for the most part, 
induced at specific places in the genetic structure. 

Since the proportion of large aberrations and non-reverting mutants is notably 
smaller among the 5-bromouracil-induced mutations, the induced changes in the 
genetic structure appear to involve small molecular substitutions rather than large 
changes of the genome. Further, the more homogeneous properties of the induced 
mutants with respect to reversion rates indicate that a certain class of molecular 
transitions may be involved here. 

Under the conditions used, the total frec^uency of occurrence of all rll mutants in 
5-bromouracil was raised several hundred fold above the spontaneous rate. The 
increase in rate at specific locations in the genetic structure was much larger. For 
example, 1 1 occurrences of mutation at one site were observed out of 67 induced rll 
mutants, while none occurred at that site among twice as many spontaneous rll 

225 



118 BIOCHEMISTRY: BENZER AND FREESE Proc. N. A. S. 

mutants. Thus the probability of mutation at that point, per repHcation of the 
phage, was raised by a factor greater than twice eleven times several hundred, or 
of the order of 10^. 

Conversely, note the site at which there were 19 occurrences among 132 spon- 
taneous r-II mutants, while among the 67 induced mutants none were observed. 
Therefore, 5-bromouracil had little, if any, positive mutagenic effect at that point. 
The disappearance of the spontaneous "hot spot" does not, of course, mean that 
spontaneous mutations at that point were suppressed but simply that they were not 
increased in proportion to other mutations. 

Only three sites (indicated by dotted lines in the figure) are represented in both 
sets of mutants. In each of those cases, there was only a single occurrence in one 
of the sets, so that the ratio of induced to spontaneous frecjuencies cannot be re- 
liably computed. The mutation rate seems to have been truly increased by 5- 
bromouracil in 2 of the cases (2 and 7 occurrences in the induced set) because one 
occurrence among the induced series represents a thousandfold greater probability 
of mutation. In the third case, there is, in the induced set, a single occurrence of a 
highly revertable mutant that appearaed nine times in the spontaneous set. This 
single event probably belongs to the small background of spontaneous mutations 
among the induced ones {ca 1 or 2 in 100). 

The methods used in the isolation of the two sets of mutants were quite different, 
and one cannot be sure whether factors other than the addition of 5-bromouracil 
could have affected the results. Although sulfanilamide alone does not appreciably 
increase the total mutation rate, the possibility that the distribution of mutations 
might change has not been excluded. Sulfanilamide, as an analogue of p-amino 
benzoic acid, inhibits the formation of folic acid, thereby inhibiting nearly all the 
methylation and hydroxymethylation steps (Cohen and Earner). i« In our sul- 
fanilamide medium, nearly all the major chemicals containing the methyl or hy- 
droxymethyl group are added except the deoxyribonucleotides of thymine and 5- 
hydroxymethylcytosine, which thus are expected to be deficient. The deficiency 
in thymine facilitates the incorporation of 5-bromouracil into DNA. Whether 
the deficiency in 5-hydroxymethylcytosine enhances the probability of the false 
incorporation of 5-bromouracil into a hydroxymethylcytosine site of the phage DNA 
remains to be seen. 

One is not in a position, from these experiments alone, to reach a clear conclusion 
as to the molecular mechanism of mutagenesis; thus, at this stage, it cannot be de- 
cided whether or not the tautomeric shift of the DNA bases from the keto- to the 
enol- form is mainly responsible for the production of point mutations, as suggested 
by Watson and Crick. ^ If it is assumed that only two kinds of nucleotide pairs are 
present in DNA (i.e., adenine— thymine, and guanine— 5-hydroxymethylcytosine). 
the existance of hot spots of spontaneous mutation and the appearance of different 
hot spots with 5-bromouracil would suggest that not every nucleotide pair of a 
given type mutates with the same probability. Rather, the mutability of a nu- 
cleotide pair would have to depend upon its position. 

In any case, the striking results of this preliminary investigation indicate that it 
would be fruitful to pursue this line of investigation, using a range of mutagenic 
substances in systems where the chemical events are under proper control. 

226 



Vol. 44, 1958 BIOCHEMISTRY: BENZER AND FREESE 119 

SUMMARY 

A set of spontaneously arising rll mutants of phage T4 is compared with a set of 
mutants induced by the action of 5-bromouracil. When analyzed by genetic 
mapping technic}ues of high resolution, the two sets of mutants are found to be 
quite different. The mutagen does not merely enhance the over-all mutation rate 
but acts at specific locations in the hereditary structure. 

The induced mutants are mostly of the nature of small, revertible alterations 
rather than gross defects. The reversion rates of induced mutants are less varied 
than those of spontaneous mutants, indicating that a certain class of molecular 
transition is involved. 

These preliminary results encourage the hope that this sort of genetic analysis 
can lead toward an understanding of the mechanism of mutation and the identifica- 
tion of the specific chemical configurations composing the genetic structure. 

Note: The occurrence at a given location of one mutation in the induced set 
corresponds to roughly 10^ limes higher mulahility of this spot in 5-bromouracil than 
is indicated by the occurrence of one mutation in the spontaneous set. 

* Aided by grants from the National Science Foundation and the American Cancer Society 

t Present address: Medical Research Council for Molecular Biology, Cavendish Laboratorj^, 
Cambridge University, England. 

X Research Fellow of the Damon Runyon Memorial Fund for Cancer Research. Present 
address: Biological Laboratories, Harvard University, Cambridge, Massachusetts. 

' J. D. Watson, and F. H. C. Crick, Cold Spring Harbor Symposia Quant. Biol, 18, 123, 1953. 

2 S. Benzer, in The Chemical Basis of Heredity, ed. McElroy and Glass (Baltimore: Johns Hop- 
kins Press, 1957). 

3 H. J. Muller, Science, 66, 84, 1927. 
' L. Stadler, Science, 68, 1928. 

5 C. Auerbach and J. M. Robson, Nature, 157, 302, 1946. 
« F. Oehlkers, Z. Ind. Abst. u. VererbungsL, 81, 313, 1943. 
' M. Demerec, Caryologia, Vol. suppl. 201, 1954. 

* M. Westergaard, Experientia, 13, 224, 1957. 
» R. Latarjet, Conipt. rend., 228, 1345, 1948. 

'» J. J. Weigle, these Proceedings, 39, 628, 1953. 

" L. Silvestri, Bull. I.S.M., 28, 193, 1949. 

'2 B. Fernandez, F. L. Haas, and O. Wyss, these Proceddings, 39, 1052, 1953. 

'3R. de Mars, Nature, 172, 964, 1953. 

'^ D. B. Dunn and J. D. Smith, Nature, 174, 304, 1954. 

'5 R. M. Litman, and A. B. Pardee, Nature, 178, 529, 1956. 

16 R. M. Herriott, and J. L. Barlow, J. Gen. Physiol, 36, 17, 1952. 

" S. Benzer, these Proceedings, 41, 344, 1955. 

'* In order to minimize the background of r mutants present in the stock of standard type, which 
is usually around 2 X 10 ~^, several stocks of T4 were grown, and one with a small proportion of r 
mutants was selected. In the stock used, this proportion was less than 10 ~^. 

15 S. S. Cohen and H. D. Barner, J. Bacterial, 71, 588, 1956. 



227 



EXPERIMENTS ON PHOTOREACTIVATION OF BACTERIOPHAGES 
INACTIVATED WITH ULTRAVIOLET RADIATION^ 

R. DULBECCO 

Department of Bacteriology, Indiana University, Bloomington, Indiana'^ 

Received for publication October 24, 1949 

Kelner (1949), working \\\ih. conidia of Streptoniyccs griseus, discovered that 
light belonging to the visible range is capable of reactivating biological ma- 
terial that has been rendered inactive by ultraviolet radiation (UV). Shortly 
after Reiner's discovery was known, a similar phenomenon in bacteriophages 
(bacterial viruses) was observed by accident. Plates of nutrient agar containing 
UV-inactivated phage and sensitive bacteria had been left for several hours on 
a table illuminated by a fluorescent lamp. After incubation it was noticed that 
the number of plaques was higher on these plates than on similar plates incu- 
bated in darkness. A short report of this phenomenon of "photoreactivation" 
(PHTR) has already been published (Dulbecco, 1949). The present paper con- 
tains the results of a first group of experiments concerning PHTR of seven bac- 
teriophages of the T group active on Escherichia coli, strain B. 

MATERIALS AND METHODS 

Stocks of each phage were prepared by inoculating material from a single 
plaque into a culture of E. coli B in a synthetic medium M9,^ except for phage 
T5, of which a stock in Difco nutrient broth was used. In some experiments the 
phage was purified by two or three steps of differential centrif ugation ; the phage 
was resuspended in m/15 phosphate buffer pH 7, with MgS04 added to a con- 
centration 10-2 M. Unless otherwise specified, the experiments described in this 
paper were performed with phage T2. Escherichia coli, strain B, was used through- 
out. In some experiments bacteria were grown in nutrient broth with aeration 
and the culture was infected with phage when it was in the logarithmic phase 
of growth (about 10^ cells per ml); these bacteria will be referred to as "bac- 
teria in broth." In other experiments bacteria were grown in broth up to a con- 
centration of about 2 X 10* cells per ml, then washed with saline (0.85 per cent 
NaCl) and resuspended in saline, kept at 37 C for 30 minutes, and then infected; 
these bacteria will be referred to as "resting bacteria." 

1 This work was done under an American Cancer Society grant, recommended by the 
Committee on Growth of the National Research Council, under the direction of Dr. S. E. 
Luria. The author wishes to express his appreciation to Dr. Luria for facilitating this work 
materially and for numerous discussions during its progress. The manuscript was completed 
at the California Institute of Technology. The author also wishes to acknowledge his in- 
debtedness to Dr. M. Delbriick for helpful discussions on the interpretation of the data. 

2 Present address : Kerckhoff Laboratories of Biology, California Institute of Technology, 
Pasadena 4, California. 

3 NH4C1,1.0 g; KH2PO4, 3.0 g; NaoHPO4,6.0 g; NaCl, 0.5 g; MgS04, 0.1 g; distilled water, 
1,000 ml; 4 g per liter glucose added after separate sterilization. 

Reprinted by permission of the author and the Williams and 

Wilkins Co., from the Journal of Bacteriology, 59 (3), 329-347, 

March, 1950. 

228 



330 R. DULBECCO [voL. 59 

Inactivation of the phages was accompHshed with a low-pressure mercury 
discharge lamp (General Electric "germicidal" lamp, 15 watts), giving most of 
the UV energy in the line 2,537 A. The output of the lamp was kept constant 
by alimenting it through a "sola" stabilizer and by using it only after it had 
been burning for at least 20 minutes. 

The stocks to be irradiated were diluted in phosphate buffer plus MgS04 and 
exposed to the lamp at a 20-inch distance either in an open petri dish with con- 
tinuous shaking (3 ml of phage in a 10-cm petri dish) or in a quartz cell 2 mm 
thick with parallel faces. Relative measurements of the incident UV doses were 
made in some experiments by timing the exposure; in other experiments rela- 



Wai-er 
Bai-h 




^=^ 



Ligh-h 
Filier 




Condenser 



[^ J H-5 Lamp 

Figure 1. Diagram of the apparatus employed for illumination in liquid. 

tive and absolute measurements were conducted with a calibrated Westinghouse 
SM-200 meter with tantalum photocell WL-775. A dose of UV will be expressed 
as seconds of exposure to the germicidal lamp. The reactivating light was used 
in two different ways: 

Illumination on the plate. The plates, prepared by the agar layer method 
(Gratia, 1936; Hershey et at., 1943), were exposed right side up to the light of 
two parallel fluorescent discharge lamps, 40 watts each, at a distance of 12 
inches at room temperature. 

Illumination in liquid. The apparatus used is illustrated in figure 1. A mercury 
discharge lamp, medium pressure (General Electric H-5 lamp, 250 watts) was 
used as the light source. The light was condensed through a spherical pyrex flask 



229 



1950] PHOTOKE ACTIVATION OF INACTIVATED BACTERIOPHAGES 331 

filled with distilled water and passed through suitable filters (see later section); 
for white light experiments infrared rays were absorbed by a filter of CuS04- 
5H2O (5 per cent in water, 1 inch thick) and ultraviolet rays shorter than 330 
m/i by a Corning glass filter no. 738. A mixture of phage and bacteria was ex- 
posed to light in a small beaker (5 ml of mixture in a beaker 4 cm in diameter) 
kept in a thermostatically regulated water bath and shaken by a reciprocating 
motion in a horizontal plane to ensure uniform distribution of the material and 
uniform illumination. Some experiments were done with a 100-watt General 
Electric Ii-4 lamp without a condenser. 

In the experiments with illumination in liquid the ratio "phage particles : bac- 
teria" was kept very low (about 10"^) to decrease the probability of multiple 
infection of bacteria and the occurrence of reactivation by multiplicity (Luria, 
1947). 

EXPERIMENTAL RESULTS 

Role of the Bacteria in the PHTR of Inactive Phage 

Phage particles inactivated by UV (UVP) can be reactivated by light only if 
the particles are mixed with sensitive bacteria during illumination. Illumination 
of UVP alone is without effect, as is shown by the following experiment : Phage 
T2 was irradiated with the germicidal lamp for 30 seconds (dark survival = 
2 X 10"^) and divided into two equal samples. The first sample was immediately 
plated and incubated in darkness; the second one was exposed to the light of a 
fluorescent lamp (80 watts at a 12-inch distance) for 1 hour at room temper- 
ature and then divided into two parts, one of which was plated and incubated 
in darkness, the other under the same light. The sum of plaque counts of two 
plates for each sample are given in table 1 (I). 

In another similar experiment the UVP was first spread on the surface of a 
nutrient agar plate and then exposed to the light; after illumination sensitive 
bacteria were spread on the same plate in darkness. In this condition also PHTR 
was not produced. 

These experiments clearly indicate that illumination of UVP in the absence 
of bacteria has no reactivating effect; they do not show, however, whether PHTR 
occurs only for adsorbed phage or also for nonadsorbed phage in the presence of 
bacteria. This point was investigated by mixing UVP with bacteria in nutrient 
broth without added NaCl (under these conditions the adsorption is slight), 
illuminating the mixture, and testing for reactivation of the nonadsorbed phage 
particles. A sample of phage irradiated with the germicidal lamp for 30 seconds 
was mixed with a culture of bacteria in broth without added NaCl, containing 
10^ cells per ml. The mixture was exposed to the light of an H-4 lamp at a 6- 
inch distance for 10 minutes at 28 C, then centrifuged; samples from the super- 
natant were plated and incubated both in darkness and in the light. The plaque 
counts (two plates for each sample) are given in table 1 (II), together with an 
assay of the irradiated phage diluted in broth by a factor equal to the one used 
in the experiment. The result of this experiment clearly indicates that the un- 
adsorbed phage particles are not reactivated by light. 



230 



332 



R. DULBECCO 



[vol. 59 



Illumination of bacteria alone followed by the addition of UVP does not pro- 
duce any PHTR. Bacteria spread on the surface of several nutrient-agar plates 
were exposed to the light of a fluorescent lamp (80 watts, 12-inch distance) for 
4 hours at room temperature ; then UVP were spread on the same plates in dark- 
ness, and the plates were incubated in darkness. Control plates, spread with 
bacteria at the same time, were kept in darkness and received UVP at the same 
time as the illuminated plates. Equal numbers of plaques were found in all plates 
whether the bacteria had been preilluminated or not, showing that preillumina- 
tion of bacteria does not cause PHTR of UVP added later. In another experi- 
ment a suspension of resting bacteria was illuminated with a light of 365-m/i 
wave length at 37 C for a period long enough to give a very high PHTR in ad- 

TABLE 1 

Ejfect of light on inactivated phage T2 alone and on unadsorhed inactivated phage T2r mixed 

with sensitive bacteria 



EXPERIMENT 


THEATMKNT 


PLAQUE 
COUNT 
0.1 ML 


I. Illumination of UVT alone 


1. UVP not illuminated and plated with 
B. Plates incubated in darkness. 

2. UVP illuminated alone and plated 
with B. Plates incubated in darkness. 

3. UVP illuminated alone and plated 
with B. Plates incubated under light. 


17 
6 

609 


II. Effect of light on unadsorbed 
phage in presence of bacteria 


1. UVP alone. 

2. UVP mixed with B in saltless broth; 
illuminated 10 minutes; centrifuged; 
supernatant plated with B; plates in- 
cubated in darkness. 

3. Same as II, 2, but with plates in- 
cubated under light. 


72 
86 

> 1,000 



sorbed UVP, and the UVP was added at the very moment at which the light 
was turned off; no measurable PHTR was observed. 

If bacteria killed by heating to 60 C for 20 minutes are substituted for living 
bacteria, no PHTR takes place. Actually the plaque count decreases, probably 
because of an irreversible adsorption of phage by the dead bacteria without the 
release of new phage. 

Illumination of bacteria prior to infection does not diminish the photoreac- 
tivability of UVP added later, as shown by the following experiment: Bacteria 
were spread on the surface of nutrient agar plates and exposed to the light of a 
fluorescent lamp (80 watts, 12-inch distance) for 4 hours at room temperature; 
then UVP was spread on the plates, which were afterwards incubated under the 
same light. After incubation the plates showed the same number of plaques as 
control plates containing nonpreilluminated bacteria and UVP, incubated under 
the same light. 



231 



1950] PHOTOREACTIVATION OF INACTIVATED BACTERIOPHAGES 333 

From these experiments with phage T2 one may conclude that PHTR occurs 
only for UVP adsorbed on sensitive bacteria and that illumination either of 
UVP or of bacteria before infection has no detectable effect. 

To test how soon after phage adsorption PHTR can occur, UVP and bacteria 
were mixed on several plates, and the plates were immediately exposed to the 
light of an H-4 lamp at an 8-inch distance at 28 C. The exposure was continued 
for 10, 20, 30, or 50 seconds. The plaque count was found to increase even after 
10 seconds, showing that no measurable delay exists between adsorption and the 
beginning of PHTR and that PHTR has no measurable latent period. 

Action of Bacterial Extracts on PHTR 

Some attempts were made to obtain PHTR by illuminating mixtures of UVP 
with cell-free bacterial extracts. Bacteria were grown in nutrient broth to a 
concentration of about 5 X 10^ cells per ml and harvested in a Sharpies centri- 
fuge. Two extraction procedures were used: (a) the thick bacterial suspension 
was frozen at —30 C, the frozen paste was then ground with carborundum pow- 
der and extracted with phosphate buffer (pH 7.5) for about 10 minutes, and the 
extract was clarified by centrifugation ; (b) the bacteria were broken in a sonic 
vibrator after the bacterial paste was diluted with an equal volume of phosphate 
buffer, and the extract was clarified in the centrifuge. In both cases the super- 
natant was a thick, yellowish liquid, which showed a high degree of enzymatic 
activity (methylene blue reduction, tryptophanase). Both extracts still con- 
tained a few living cells, which could be eliminated either by filtration or by 
repeated freezing at —30 C. 

UVP was mixed into various dilutions of the extracts, and the mixtures were 
kept either in light or darkness and assayed for active phages at different times. 
Only extracts still containing living cells gave some PHTR. Removal of almost 
all living cells eliminated PHTR. 

PHTR as a Function of the Dose of the Inactivating UV Light 

For several phages of the T group (T2, T4, T5, T6) the curve obtained by 
plotting the logarithm of the active fraction against the UV dose approaches 
a straight line (Latarjet and Wahl, 1945), at least for high values of the dose, 
whereas an inflection with downward concavity, of dubious origin, may appear 
for low doses. Three other phages (Tl, T3, T7) show, on the contrary, an inflec- 
tion with upward concavity of unknown origin. 

If the inactivated phages are adsorbed on bacteria and exposed to light of 
high intensity for a sufficient length of time, the active fraction increases and 
reaches a maximum (see later section). After this maximum is reached, the 
curve showing the logarithm of the active fraction against the UV dose has for 
each phage the same shape as the curve obtained in darkness, but for a given 
UV dose the slope of the curve obtained after PHTR is lower than the slope of 
the curve in darkness. 

The fact that both curves in the light and in darkness tend to be straight 
lines with different slopes for high UV doses is an indication that absorption of 



232 



334 



R. DULBECCO 



[vol. 59 



UV light in the phage has a probability, a, of producing a photoreactivable in- 
activation and a probability, b, of producing a nonphotoreactivable inactivation; 
a + 6, the probability of producing any inactivating damage, is proportional 
to the cross section of the phage for UV. Assuming a -{- b = 1, a is the photo- 
reactivable sector of the cross section, b the nonphotoreactivable sector; h is 
measured by the ratio of the slope of the curve after maximum PHTR to the 
slope of the curve in darkness, both measured in the straight parts. 

The photoreactivable sector, a, varies between 1 (complete photoreactivability) 
and (no photoreactivability) and can therefore be used as an index of the 
photoreactivability. Values of a for different phages are given in table 2. 

Influence on PHTR of the Interval of Time between Infection and Exposure to Light 

In the experiments reported in the present and following sections the influence 
of various experimental conditions on PHTR was analyzed. A quantitative de- 
termination of PHTR was made by measuring either the "active fraction" or 
the "amount of PHTR" in an UVP sample after a given exposure to light. The 
active fraction is the ratio of the number of active particles after PHTR to the 
total number of adsorbed particles and is equal to the sum of the fraction active 

TABLE 2 
Photoreactivability of the phages of the T group 



PHAGE 


Tl 


T2 


T3 


T4 


T5 


T6 


T7 


Photoreactivable sector of cross section 
(a) 


0.68 


0.56 


0.39 


0.20 


0.20 


0.44 


0.35 



in the darkness plus the fraction reactivated by light; the amount of PHTR is 
the reactivated fraction. 

The influence on PHTR of the interval of time between infection and exposure 
to light was determined for UVP adsorbed on bacteria in broth and on resting 
bacteria (see "Material and Methods"). Bacteria and UVP were mixed in dark- 
ness; samples of the mixture were kept in darkness for various intervals of time 
and then exposed to light for a period long enough to produce maximum PHTR. 
After illumination, samples were plated and incubated in darkness, and the ac- 
tive fraction was determined. In this procedure the bacteria infected with ir- 
radiated phage particles had to be exposed to light much longer than the latent 
period between infection and liberation of phage adsorbed on bacteria in broth. 
When bacteria in broth were used, therefore, the mixtures were plated before the 
end of the latent period and illumination was continued by exposing the plates; 
when resting bacteria were used, illumination could be continued indefinitely in 
liquid, since no phage liberation takes place under these conditions. 

Experiments with bacteria in broth. The experiments were performed with phage 
T2 at 28 C. The amount of PHTR decreased rapidly as the time interval between 
infection and the beginning of exposure to light increased; after about 20 minutes 
only a small amount of PHTR was produced, as is shown in table 3. 



233 



1950] PHOTOREACTIVATION OF INACTIVATED BACTERIOPHAGES 335 

This decrease in PHTR might be caused by a gradual decrease in the amount 
of PHTR per time unit as the time interval between infection and illumination 
increases, by a limitation of the time interval after infection in which PHTR 
can occur, or by both. The amount of PHTR per time unit was determined in 
experiments in which exposure to light was started at various times after in- 
fection. The results, shown in figure 2, indicate that the amount of PHTR per 
time unit remained practically constant for about 15 minutes. The decline in 
maximum PHTR must be due, therefore, to a limitation of the time within which 
PHTR can occur after infection, the useful time interval ending between 20 and 
30 minutes after infection under the experimental conditions; after this time very 
little or no PHTR can take place. 

Experiments with resting bacteria. As is shown in table 4, the maximum amount 
of PHTR obtainable in phage T2r irradiated with the germicidal lamp for 18 
seconds remains fairly constant for at least 70 minutes after infection at 37 C; 

TABLE 3 
The effect of the time interval between infection and exposure to light (bacteria in broth) 
Phage T2r, irradiated for 20 seconds with the germicidal lamp, was mixed with bacteria 
and adsorption was allowed to continued for 2 minutes, after which it was interrupted by 
serum anti-T2. Exposure to light (H-4 lamp, 12-inch distance) was begun at various times 
and was continued for 100 minutes at 28 C. Amount of PHTR is lower than in experiment 
reported in table 4, because in the present experiment a lower light intensity was used, 
and the time in which the light could be utilized for reactivation was limited, since bacteria 
in broth were used. 



TIME INTERVAL BETWEEN INFECTION AND EXPOSURE TO UGHT 


ACTIVE FRACTION 


min 







5.3 X 10-5 


10 


1.4 X 10-3 


20 


3.5 X lO-" 


30 


5.0 X lO-" 


Active fraction in darkness 


3.0 X 10-" 



longer intervals have not been tested. The amount of PHTR per time unit is 
not influenced by the time interval between infection and illumination. 

The differences between experiments with bacteria in broth and with resting 
bacteria indicate that under the experimental conditions the system "UVP- 
metabolizing bacteria" undergoes a gradual change that in its late phases pre- 
vents PHTR, a change absent in the system "UVP-resting bacteria." 

Kinetics of PHTR 

PHTR as a function of the time of exposure to the reactivating light. The following 
experiments employed inactive phage T2r and resting bacteria, with illumina- 
tion in liquid. Inactive phage diluted in phosphate buffer was mixed with bac- 
teria at time at 37 C in darkness, and 10 minutes were allowed for complete 
adsorption. At the eleventh minute a sample was plated in darkness; at the 
twelfth minute the mixture was exposed to light, and samples were taken at 



234 



336 



R. DULBECCO 



[vol. 59 




; X /o 1- 



10 15 

t Time (rninu-hes) 

Figure 2. The fraction of active particles as a function of the time of illumination (in 
minutes) and of the interval between infection and exposure to light. Each curve gives the 
active fraction as a function of the time of illumination (in minutes) for a different interval 
between infection and exposure to light; the interval is indicated (in minutes) at the right 
end of each curve. Phage T2r was irradiated for 20 seconds with the germicidal lamp, ad- 
sorbed on bacteria in broth, and exposed to light in broth at 28 C. 



TABLE 4 

The effect oj the time interval between infection and exposure to light {resting bacteria) 
Phage T2r, irradiated for 18 seconds with the germicidal lamp, was adsorbed onto resting 
bacteria suspended in saline. Exposure to light (H-5 lamp with condenser, wave length 365 
mn) began at various times. Illumination was carried out in liquid. 



TIME INTERVAL BETWEEN INFECTION AND EXPOSURE TO UGHT 


ACTIVE FRACTION 


min 







5 X 10-2 


10 


5 X 10-^ 


30 


6 X 10-2 


50 


5.4 X 10-2 


70 


6 X 10-2 


Active fraction in darkness 


10-3 



various time intervals thereafter and plated in darkness; all plates were incu- 
bated in darkness. The experiments lasted 140 minutes at most; control experi 



235 



1950] 



PHOTOREACTIVATION OF INACTIVATED BACTERIOPHAGES 



337 



merits showed that resting bacteria that have adsorbed active phage do not 
Uberate any phage in this time interval. The active fraction always increased 
with the time of illumination, the increase becoming less and less with increasing 
time, so that a maximum was reached as is shown in figure 3. The time at which 
the maximum was reached depended on the light intensity, a longer time being 
required when the intensity was lower; when the light intensity was varied in 
such a way that the maximum was reached in a period between 20 and 140 
minutes, approximately the same maximum was reached in all cases, as is shown 
in figure 3. 



6 y 10 



5 xlO' 



.-2 



n r 



"I I I r 




ZO 30 40 50 60 70 80 90 

Exposure to fhe Reaciivafin^ Ligh-h (minufes) 

Figure 3. The fraction of active particles as a function of the time of illumination and of 
the light intensity. The active fraction is plotted against the time of illumination (in 
minutes). Phage T2r was irradiated for 20 seconds with the germicidal lamp, adsorbed 
on resting bacteria, and illuminated in liquid at 37 C. Curve 1 was obtained with a light 
of intensity 10 (in arbitrary units), curve 2 with a light of intensity 2.9, and curve 3 with 
a light of intensity 0.6. 

The amount of PHTR (defined in previous section) observed in a sample of 
UVP after a given time of illumination (pit)), divided by the amount of PHTR 
obtained in the same sample when PHTR has reached the maximum value (p 
(qo)), will be indicated at the "relative amount of PHTR"; it can vary between 
zero and one. 

By subtracting the relative amount of PHTR from unity, one obtains the 
fraction of photoreactivable particles that are still inactive after a time, t, of 

illumination ( = 1 — — - — : 1 . The logarithm of this quantity plotted against 

the time of illumination always gave a straight line for different intensities of 



236 



338 



R. DULBECCO 



[vol. 59 



the reactivating light and for different doses of the inactivating UV. A curve of 
this type is reproduced in figure 4. The hnearity of the experimental curves was 
found to be statistically significant by comparing, with the x" test, the experi- 
mental data for the active fractions with data calculated on this assumption, as 
is shown in table 5. This result shows that PHTR is a one-hit phenomenon; a 




40 &0 

Time of IHuminafion ^minufes^ 

Figure 4. The logarithm of the fraction of photoreactivable particles that has not been 

reactivated after a given time of illumination ( 1 — -— — r ) plotted against the time of illum- 

\ P(«) / 
ination (in minutes). Phage T2r was irradiated for 20 seconds with the germicidal lamp, 
adsorbed on resting bacteria, and illuminated in liquid at 37 C. 

TABLE 5 

A comparison between observed and calculated active fractions after different times of 

illumination 



KXPKRIMENT NO. 


UV DOSE IN SECONDS 


DEGREES OF FREEDOM 


x' 


p 


312 


10 


10 


12.1 


>0.20 


313 


30 


11 


10.6 


>0.40 


315 


20 


12 


14.3 


>0.20 


319 


20 


11 


19.8 


0.05 



photoreactivable particle is reactivated by one quantum only, independently of 
the UV dose. 

The dependence of the amount of PHTR on the time of illumination is ex- 
pressed by the equation 

p(t) = (1 - e-n F(D) 

in which t is the time of illumination, F{D) the photoreactivable fraction, which 
is a function of the dose, D, of UV. Value / is the probability per time unit that 



237 



1950] PHOTOKE ACTIVATION OF INACTIVATED BACTERIOPHAGES 339 

a particle is photoreactivated and can be called the PHTR rate; it is proportional 

to the slope of the line giving log f 1 — —, — : j versus time of illumination. Value 

/ may depend on several variables, such as dose of UV, intensity of the reactivat- 
ing light, temperature, and metabolic condition of the bacteria during PHTR. 
This dependence will be examined in the next sections. 

Dependence of PHTR rate on dose of UV and intensity of the reactivating light. 
PHTR rate (/) was determined for UVP inactivated with different UV doses, 
adsorbed on resting bacteria, and illuminated in liquid at 37 C with light of 
constant intensity; it was found to be approximately constant for doses of UV 
between 10 and 30 seconds. The results, however, are not yet definite on this 
point, and a decrease of / by a factor 1.2 when the UV dose increases from 10 
to 30 seconds cannot be excluded. This result shows that the probability for an 
adsorbed quantum to reactivate a photoreactivable phage particle is practically 
independent of the inactivating UV dose. 

Value / was also determined for different intensities of the reactivating light 
on UVP inactivated with the same UV dose, adsorbed on resting bacteria, and 
illuminated in liquid at constant temperature. The intensity was varied either 
by changing the distance of the sample from an H-4 lamp— assuming the in- 
tensity to be inversely proportional to the square of the distance — or by lowering 
the intensity of a monochromatic light by filters and measuring with a thermo- 
pile the relative intensities. Value / was found to increase almost linearly with 
light intensity for low intensities but for high intensities to tend to a maximum 
as is shown in figure 5. The highest value of the PHTR rate observed in these 
experiments was about 1.4 X 10"^ sec"^ and corresponds to a half-time of about 
8 minutes. 

For low light intensities, / being a linear function of the intensity, the proba- 
bility of PHTR occurring in a bacterium-phage complex is a linear function of 
the dose of the reactivating light (equal to intensity X time), whereas for high 
intensities the same dose has less effect. For low intensities and relatively short 
exposures the dependence of amount of PHTR on light dose is also approxi- 
mately linear. 

Action Spectrum of PHTR 

Seven wave lengths were tested for photoreactivating activity. The corre- 
sponding monochromatic lights were obtained in the following ways (see Bowen, 
1946): 

(1) Group of lines near 313 m^t of the mercury arc (with a small amount of 
334 mju). Light: mercury lamp H-4 without glass envelope. Filter: 3 cm NiSOi- 
7H2O, 350 g + CoS04-7H20, 10 g made up to a liter with water; 1 cm potas- 
sium hydrogen phthalate, 5 g in 1,000 ml water. 

(2) Group of lines 365 m^t of the mercury arc. Mercury lamp H-5 (General 
Electric) ; Coming glass filter combination nos. 738, 5860. 

(3) Group of lines 404 uifi of the mercury arc. Lamp H-5. Filter: 2 
cm Cu(N03)o-6H20, 200 g in 100 ml water. Iodine 0.75 g in 100 ml carbon tet- 
rachloride. 



238 



340 



R. DULBECCO 



[vol. 59 



(4) Group of lines 434 irijLi of the mercury arc. Lamp H-5. Filter: 2 cm CuS04- 
5H2O, 25 g + 300 ml ammonium hydroxide (d = 0.88), made up to 1 liter 
with water; 1 cm NaNOo, 75 g in 100 ml water. 

(5) Band around 500 m^u (between 480 and 520 myu, center 500 m/z). Pro- 
jection lamp with ribbon filament. Filter: Wratten no. 47, Wratten no. 58, 2 cm 
CUSO4, 5 per cent. 

(6) Line 546 m/x of the mercury arc. Mercury discharge lamp H-5. Corning 
glass filter combination, nos. 3484, 4303, 5120. 



/•» xlO 


I 


- 1- -T" — - -1 1 1 


^J^ — ' ' " 


IZ%IO'^ 


- 


Ox-^''^ 


- 


1 X /O'-' 




x^ 


- 


^a X 10" 


0/ 




- 


6 X 10'' 


- jo 




- 


4 X /O"" 


\ 




- 


2 X 10'" 


1 




r 1 . 



Relaf-ive Lighi- In+ensH-y (arbii-rary unH-sy 

Figure 5. The PHTR rate as a function of the intensity of the reactivating light. The 
PHTR rate is expressed in sec~^ and the light intensity in arbitrary units. Phage T2r was 
irradiated for 20 seconds with the germicidal lamp, adsorbed on resting bacteria, and il- 
luminated in liquid at 37 C. 

(7) Group 576-579 m/n of the mercury arc. Lamp H-5. Filter: 1 cm mixture 
of CuCl2-2H20, 10 g in 10 ml water and CaCls, 3 m, 90 ml; 2 cm KoCrjO?, 15 g 
in 200 ml water. 

The efficiency of the different lights was determined in the following way: 
For each light the range of intensity w'as first determined, in which the PHTR 
rate is approximately proportional to the light intensity; the intensity of the 
most effective wave lengths was reduced by filters until it fell into this range. 
The rate of PHTR was then determined for each wave length, and the relative 
intensities were measured with a thermopile. The time of illumination was short, 
so that the amount of PHTR was very nearly proportional to the dose of reac- 
tivating light (see previous section). 

The dose of light of each wave length necessary to give a standard amount of 



239 



1950] 



PHOTOREACTIVATION OF INACTIVATED BACTERIOPHAGES 



341 



PHTR in a given time was calculated from these data, and the reciprocal of this 
dose (given in arbitrary units) was taken as a measure of the activity of that 
wave length. In figure 6 the activity of the wave lengths tested is plotted against 
the wave length. 

The activity of a given light may be underestimated, since it is known that 
light of the wave lengths used in PHTR may damage the bacteria (Hollaender, 
1943) or the phages (Wahl and Latarjet, 1947). The killing action of the seven 
wave lengths on active phage adsorbed on bacteria was therefore determined, 
and it was found that with the light intensity and the time of illumination used 
in the PHTR experiments an appreciable killing activity was only evident for 
wave length 313 m^. To correct for this killing activity, the amount of PHTR 




3 4000 5000 6000 

yVoi'e Lenq-i-h in A 

Figure 6. The action spectrum of PHTR. The activity of each wave length, given in 
arbitrary units, is plotted against the wave length. Phage T2r was irradiated for 20 seconds 
with the germicidal lamp, adsorbed on resting bacteria, and illuminated in liquid at 37 C. 

obtained after a given exposure to this light was increased bj^ a factor equal to 
the decrease in titer of active phage adsorbed on B exposed to the same light 
for the same length of time in equal experimental conditions. The curve of PHTR 
as a function of the time of exposure to 313 m/z light, obtained in this way, was 
almost linear and was used in calculating the activity of the light. 

The activity of the seven wave lengths tested gives only the general shape of 
the action spectrum. It consists of a band covering the range from about 300 ran 
on the side of the short wave lengths to about 500 m/x on the side of the long 
ones, with a maximum around 365 m/x. The greatest photoreactivating activity 
occurs therefore in the near ultraviolet. 

The action spectrum of PHTR is related to the absorption spectrum of the 



240 



342 



R. DULBECCO 



[vol. 59 



pigment that absorbs the reactivating hght (see Loofbourow, 1948, for dis- 
cussion of this relation); we tried, therefore, to obtain on this basis some infor- 
mation about the photosensitive pigment. The action spectrum is not detailed 
enough to give a specific indication; it shows, however, that the pigment is not 
contained in the unmodified phage, since the absorption spectrum of purified 





■\ 1 


■■| 


B 


.500 


/ 


\ .300 










\ 


.4O0 


\j 


\ .ZOO 


\ 


^.300 
c 
Q 




9 .100 


0..~.O. 0— 0.~S3 

1 t 1 1 1 1 1 


.200 






320 360 400 440 




■ 






.100 


A . . 




■ 

1 1 r r 1 1 1 1 



320 340 360 380 

Wave Len^i-h J mu 



440 



Figure 7 . Absorption spectra of purified phage T2. The optical density is plotted against 
the wave length. The spectra were obtained with a Beckman quartz spectrophotometer. 
The phage was suspended in phosphate buffer (m/15, pH 7) plus MgS04 10"' u. A. Active 
phage, concentration 5.5 X 10^" infecting units per ml. B. Upper curve : absorption spectrum 
of active phage for wave lengths longer than 320 m^t, concentration 2.3 X 10" infecting units 
per ml; lower curve: absorption spectrum of the same phage after 2 hours' irradiation with 
the germicidal lamp at a distance of 12 inches. 

phage has no band comparable to the action spectrum of PHTR. This is shown 
by figure lA, which reproduces the absorption spectrum of purified phage T2. 
For wave lengths longer than 320 mju the absorption closely follows Rayleigh's 
law of scattering and is, therefore, due to scattering of the light. This is shown 
more convincingly by plotting the logarithm of the optical density at different 
wave lengths against the logarithm of the wave length; according to Rayleigh's 



241 



1950] 



PHOTOKEACTIVATIOX OF IXACTIVATED BACTERIOPHAGES 



343 



law one should obtain a straight line with slope 4 (see Oster, 1948, 323, formula 
6). As shown in figure 8A, the curve, obtained from the same data used for figure 
7A, is a straight line in the range of wave lengths 320 to 450 m^t; the slope of 
the curve is 3.7, instead of 4, owing to the size of the phage particles, which is 
larger than required for the strict application of Rayleigh's law (La Mer, 1948). 
The pigment might be formed in the phage after UV irradiation. To check 
this point a suspension of purified phage T2 containing 2.3 X 10'^ particles per ml 
was irradiated in an open shallow container with the germicidal lamp at a dis- 
tance of 12 inches for variable lengths of time up to 4 hours, and the absorption 



1800 






/400 




loo X 

Figure 8. The logarithm of the optical density, D, of purified phage T2 versus the lo- 
garithm of the wave length for the range 320 to 450 mju. A. Active phage (the same data as 
in figure 7 A). B. Phage after long UV irradiation (same data as in figure IB). Dashed line 
shows the curve expected for pure scattering. 

spectrum was determined at regular intervals. The absorption spectrum was 
found to undergo complex changes as the irradiation proceeded; we shall limit 
our attention to the modifications occurring in the range of wave lengths longer 
than 320 m^u. A general decrease in absorption in this region was observed, and 
after about one hour of irradiation a faint band was noticed, which became more 
evident during the next hour (figure 7B). This band has maximum absorption 
around 330 m/i and extends to about 380 m/x on the side of longer wave lengths 
(figure 85) ; its limit on the side of shorter wave lengths cannot be determined 
because of overlapping with the general phage absorption. 



242 



344 



K. DULBECCO 



[vol. 59 



The maximum absorption of this band is located at a shorter wave length 
than the maximum of the action spectrum of PHTR; this difference, however, 
is not such as to exclude the band from belonging to the photosensitive pigment, 
because the location of the band may be shifted toward longer wave lengths if 
the pigment is bound with some bacterial constituent after adsorption of the 
inactive phage on bacteria. 

Influence of the Metabolic Conditions of Bacteria in PHTR of Phages 

PHTR is not appreciably different whether the phage is adsorbed by bacteria 
in broth or in a synthetic medium; the rate of PHTR is somewhat lower with 
resting bacteria than with bacteria suspended in nutrient media. 

The influence of oxygen on PHTR was determined. Resting bacteria and UVP 
were placed in separate compartments of a Thunberg tube, nitrogen was bubbled 
for about 20 minutes through the bacterial suspension, the tube was then evacu- 
ated by a pump, and air was replaced with nitrogen, the operation being re- 

TABLE 6 
PHTR rate and Qio at different temperatures 
The Qio was determined from the ratio of Qd of the rates at two successive temperatures, 
using the formula: Qio = (Qd)'"''', where d is the temperature interval between two ob- 
servations. Phage T2r irradiated for 20 seconds with the germicidal lamp, illuminated in 
liquid. 



TEMPERATURE (C) 


PHTR RATE (SEC"') 


Qio 


3 


1.25 


X 10- » 




11 


6.2 


X 10-5 


7.5 


16 


1.3 


X 10-^ 


4.2 


24 


2.3 


X 10-^ 


2.1 


30 


3.9 


X 10-* 


2.4 


37 


5.6 


X 10-* 


1.7 



peated 4 times. Phage and bacteria were mixed and the tube was exposed to 
light. The control consisted of an open tube from which oxygen had not been 
removed. The same amount of PHTR was observed in both tubes. Oxygen is 
therefore not necessary for PHTR, at least not for the initial photochemical re- 
action. We also found that cyanide in 10"^ m concentration does not affect PHTR. 

The Effect of Temperalure on PHTR 

PHTR can occur at temperatures too low to allow growth of active phage 
(+1 C). To obtain a measurable amount of PHTR at this temperature, one 
must mix UVP and bacteria in the dark at 37 C to allow enough adsorption, 
then chill the mixture to 1 C and expose it to light. 

Determinations of PHTR rate with constant illumination were made at 3, 11, 
16, 24, and 37 C, using phage T2 irradiated for 20 minutes with the germicidal 
lamp. The Qio was determined for each interval, and the results are reproduced 
in table 6. 



243 



1950] PHOTOREACTIVATION OF INACTIVATED BACTERIOPHAGES 345 

The behavior of the Qio is similar to that found for complex bacterial activities 
and for enzymatic reactions (Rahn, 1932) and is an indication that the physio- 
logical state of the bacterium conditions the probability of the photoreactivating 
event. 

PHTR of Phage Inactivated with X-Rays 

It was previously reported (1949) that no PHTR had been detected for phage 
inactivated by X-rays. This statement is valid only if X-ray irradiation is per- 
formed on phage in synthetic medium. With phage T2 inactivated by X-rays in 
nutrient broth a slight amount of PHTR can be observed. This amount is proba- 
bly reduced by the poor adsorption of p^age inactivated by X-rays, as discovered 
by Watson (1948). After correction for the limited adsorption (data kindly sup- 
plied by Watson), the PHTR of X-ray -inactivated phage is still considerably 
less than that for phage inactivated to the same extent by UV. 

SUMMARY AND DISCUSSION 

In the following brief discussion we shall try to arrange the results of our ex- 
periments in an order that will bring out their theoretical implications, and we 
shall present a working hypothesis for the mechanism of PHTR. 

(1) The damage caused by UV in bacteriophage consists of two kinds: photo- 
reactivable and nonphotoreactivable damage. These two kinds of damage occur 
with comparable cross sections; they may reflect the presence of two kinds of 
UV-absorbing constituents in each phage particle. Further information should 
be obtainable by determining the action spectrum for the two types of inac- 
tivation. 

(2) Only phage particles adsorbed on bacteria undergo PHTR; PHTR can 
occur within a few seconds after adsorption, indicating that PHTR is due to 
reactions occurring in the early phase of the interaction between inactive phage 
and bacterium; perhaps surface reactions are involved, and this may account 
for the failure to reproduce PHTR with bacterial extracts. PHTR does not re- 
quire the presence of external metabolic substrates or of oxygen and is not in- 
hibited by cyanide, but is influenced by the physiological condition of the bac- 
teria after infection. 

(3) The photosensitive pigment has an action spectrum with a maximum near 
365 m/x. Normal phage does not have an absorption band corresponding to this 
action spectrum. UV-treated phage shows an absorption band with a maximum 
near 330 m/z. Perhaps this is the photosensitive pigment created by UV irradia- 
tion in the phage. The shift from 330 to 365 m/x could be due to the binding of 
the pigment with bacterial constituents upon adsorption of a phage particle on 
the bacterium. The alternative possibility is that the photosensitive pigment 
exists in the bacterium prior to infection. Further studies of the absorption band 
of UV-irradiated phage and of the action spectrum of PHTR are needed. 

(4) At low intensities of illumination the probability of PHTR occurring in a 
bacterium-phage complex is proportional to the dose of light. From this we 
might conclude that the individual light quanta absorbed by the bacterium- 



244 



346 R. DULBECCO [voL. 59 

phage complex do not co-operate to produce PHTR but that each quantum in- 
dividually has a chance to accomplish PHTR and that this chance is independent 
of any other quanta absorbed by the complex. We would thus be led to conjec- 
ture that PHTR is due to a primary photochemical reaction. 

(5) The picture is complicated by the finding that at high intensities the rate 
of PHTR ceases to be proportional to the intensity of illumination, reaching a 
maximum value, and by the finding of a complex temperature dependence. These 
findings require the participation of dark reactions in the mechanism of PHTR. 

(6) The probability of PHTR per time unit (PHTR rate) is practically in- 
dependent of the dose of U\^ used for inactivation. This finding shows that 
photoreactivation is an all-or-none phenomenon, and it may indicate that photo- 
reactivable inactivation is always due to one injury, elimination of which resti- 
tutes activity. 

To explain all the Icnown features of PHTR, we propose the following working 
hypothesis: The photoreactivable inactivation is due to formation of molecules 
of an inhibitor in the phage; although many of these molecules may be formed 
in one phage particle, just one molecule is the inactivating one in each case, 
for example, by blocking a reaction necessary for phage growth in a small area 
of the contact surface between phage and bacterium. Restoration of phage ac- 
tivity requires the permanent removal of the inactivating inhibitor molecule. 
Dissociation does not occur by thermal activation; but absorption of a light 
quantum of a given wave length produces a transient and reversible dissociation. 
During the time the inhibitor is dissociated it can be captured by a receptor and 
destroyed with dark reaction. This makes the removal permanent and consti- 
tutes reactivation. 

PHTR therefore requires a system made by phage, inhibitor, pigment, and 
receptor. The inhibitor belongs to the phage, the receptor to the bacterium being 
perhaps of enzymatic nature ; the pigment may belong to either one and may be 
identified either with the inhibitor or with the receptor. The system is com- 
pleted after adsorption of the inactive phage on bacteria. 

The probability of PHTR (PHTR rate) for low light intensity is proportional 
to the time integral in which the inhibitor is dissociated and therefore to the 
number of quanta absorbed (dose of the reactivating light). For high intensities 
the activation times due to absorption of different quanta overlap somewhat, 
so that equal doses become less efficient. When the light intensity is so high that 
the inhibitor is dissociated without interruption during illumination, the proba- 
bility of PHTR reaches a maximum value. The probability of PHTR is also pro- 
portional to the probability that the dissociated inhibitor will be captured and 
destroyed by the receptor in the time unit and therefore depends on temper- 
ature, which influences the dark reactions, and on some physiological conditions 
of the bacteria, which may affect the efficiency or the amount of the receptor. 

REFERENCES 

BowEN, E. J. 1946 The chemical aspects of light. Clarendon Press, Oxford. 
DuLBEcco, R. 1949 Reactivation of ultraviolet-inactivated bacteriophage by visible 
light. Nature, 163, 949-950. 



245 



1950] PHOTOREACTIVATION OF INACTIVATED BACTERIOPHAGES 347 

Gratia, A. 1936 Des relations num^riques entre bact^ries lysogenes et particules de 

bacteriophage. Ann. inst. Pasteur, 57, 652-676. 
Hershey, a. D., Kalmanson, G., and Bronfenbrenner, J. 1943 Quantitative methods 

in the study of the phage-antiphage reaction. J. Immunol., 46, 267-280. 
HoLLAENDER, A. 1943 Effect of long ultraviolet and short visible radiation (3500 to 

4900A) on Escherichia coli. J. Bact., 46, 531-541. 
Kelner, A. 1949 Effect of visible light on the recovery of Streptomyces griseus conidia 

from ultra-violet irradiation injury. Proc. Natl. Acad. Sci. U. S., 35, 73-79. 
La Mer, V. K. 1948 Monodisperse colloids and higher-order Tyndall spectra. J. 

Phys. Colloid Chem., 52, 65. 
Latarjet, R., and Wahl, R. 1945 Precisions sur I'inactivation des bacteriophages par 

les rayons ultra-violets. Ann. inst. Pasteur, 71, 336-339. 
LooFBOUROw, J. R. 1948 Effects of ultraviolet radiation on cells. Growth, 12, Suppl., 

75-149. 
LuRiA, S. E. 1947 Reactivation of irradiated bacteriophage by transfer of self-reproduc- 
ing units. Proc. Natl. Acad. Sci. U. S., 33, 253-264. 
LuRiA, S. E., AND DelbrIjck, M. 1942 Interference between bacterial viruses. II. 

Interference between inactivated bacterial virus and active virus of the same strain 

and of different strain. Arch. Biochem., 1, 207-218. 
Oster, G. 1948 The scattering of light and its applications to chemistry. Chem. Rev., 

43, 319-365. 
Rahn, O. 1932 Physiology of bacteria. P. Blakiston's Son and Co., Philadelphia. 

Refer to p. 123, 221. 
Wahl, R., and Latarjet, R. 1947 Inactivation de bacteriophages par des radiations de 

grande longeur d'onde (3400-6000 A). Ann. inst. Pasteur, 73, 957-971. 
Watson, J. D. 1948 Inactivating mutations produced by X-rays in bacteriophages. 

Genetics, 33, 633. 



246 



GENETIC RECOMBINATIONS LEADING TO PRODUCTION OF 

ACTIVE BACTERIOPHAGE FROM ULTRAVIOLET 

INACTIVATED BACTERIOPHAGE PARTICLES^ 

S. E. LURIA AND R. DULBECCO 

Department oj Bacteriology, Indiana University, Bloomington, Indiana 
WITH AN APPENDIX BY R. DULBECCO 

Received June 28, 1948 

THE potentialities of bacteriophage genetics have been revealed by the 
discovery of genetic recombinations among related phage particles in- 
fecting the same bacterial cell (Delbrijck and Bailey 1946). The complexities 
of these genetic systems have been further illustrated by the work of Hershey 
and RoTMAN (1948) on a number of different genetic determinants involved 
in the determination of the alternative phenotypes r+ and r in phage T2H. 
A different approach to the genetic mechanisms of bacteriophages originated 
from the chance observation by Delbrijck and Bailey that, after exposure 
to ultraviolet light, some phages gave variable plaque counts, depending on 
the relative concentrations of phage lysates and host cells at the time of assay. 
In investigating this phenomenon, one of us (Luria 1947) discovered a mecha- 
nism of phage reactivation by interaction among inactive particles in the 
course of intracellular growth. The detailed investigation of this phenomenon 
has indicated new possibilities for a quantitative analysis of the genetic struc- 
ture of these viruses and has suggested a possible mechanism for their repro- 
duction. 

A preliminary discussion of some of the results reported in this paper has 
appeared (Luria 1947) ; their implications for a number of problems have been 
discussed in a forthcoming publication (Luria 1948). The present article is 
intended to present the results in detail, and, by describing techniques and 
methods of analysis, to serve as a background for future publications on this 
topic. 

The analysis of the results presented in the following pages is based on the 
hypothesis that inactivation of bacteriophage particles by ultraviolet light is 
due to production of discrete alterations in individual portions of genetic ma- 
terial. Although the internal evidence in support of this hypothesis, as pre- 
sented in this paper, is quite satisfactory, it must be said that satisfactory 
external evidence from other lines of attack is not yet available. The conclu- 
sions reached in this article must be considered for the time being as working 
hypotheses for further investigation. 

material and general methods 

The system of phages T1-T7, their r mutants, and their common host 
Escherichia coli strain B have repeatedly been described, as well as the use of 

' This work was done under an American Cancer Society grant recommended by the 
Committee on Growth of the National Research Council. We wish to acknowledge the able 
assistance of Mrs. J. P. Headdy. 



Reprinted by permission of the authors and Genetics, Inc., from 
Genetics, 34, 93-125, March, 1949. 

247 



94 S. E. LURIA AND R. DULBECCO 

bacterial mutants resistant to one or more phages as indicators for one phage 
in the presence of another (see Delbruck 1946). Plate counts for viable bac- 
teria, and plaque counts in agar layer for active phage were used throughout, 
employing 1.1 percent agar in "Difco" nutrient broth plus 0.5 percent NaCl. 
All plates were incubated at 37°C. Experimental bacterial cultures in the 
logarithmic phase of growth were grown with aeration at 37°C from standard 
inocula. 

The phage stocks were lysates in glucose+ammonia (or lactate+ammonia) 
medium. These media give negligible absorption of the ultraviolet light used 
in this work. High titer phage lysates (over 1X10" particles per ml) might 
give some ultraviolet screening effect because of bacterial debris and of phage 
itself. Whenever possible, therefore, phage was irradiated after a dilution 1 :5 
or higher in the same medium. The source of ultraviolet was a General Electric 
Company germicidal bulb, 15 watts, alimented through a stabilizer. At a 
distance of 50 cm from the center of this bulb, the flux— measured with a West- 
inghouse SM-200 meter with tantalum phototube WL-775 — is about 7 
ergXmm"^ sec~^ The beam contains mainly radiation of wavelength 2537 A. 
Samples were irradiated in a thin layer (not over 0.4 mm) in open Petri 
dishes rocked during exposure. 

The technique of "one-step growth" experiment in its various forms has 
been described in detail previously (Delbruck and Luria 1942). 

EXPERIMENTAL 

Inactivalion and reactivation of bacteriophages 

Plaque counts on phage suspensions exposed to ultraviolet for various lengths 
of time generally give survival ratios whose logarithms are proportional to the 
dose, that is, to the time of exposure (see Latarjet and Wahl 1945, and 
figure 1). The logarithmic rate indicates a one-hit mechanism of inactivation 
(Lea 1947), and we can assume that the hit consists of the successful absorption 
of one quantum. The probability that one quantum produces inactivation is, 
however, very small: for phage T2, for example, one inactivating hit is pro- 
duced by a dose corresponding to almost 10^ quanta absorbed per particle 
(M. Zelle, personal communication). Only one absorption in 10^ on the aver- 
age is, therefore, effective, the others probably producing excitations that do 
not lead to the inactivating effect. 

When the average number of effective hits per particle is r, the proportion 
of active to total phage will be e"*". For r = 1, e~'' = 0.37 ; the corresponding dose 
is the "inactivation dose" in Lea's terminology (1947). If doses are expressed 
in multiples of the inactivation dose, their values give directly the average 
number of hits per particle. 

Phage particles inactivated by ultraviolet light are adsorbed by bacteria 
(Luria and Delbruck 1942). This is detected because adsorption of one par- 
ticle by a bacterium causes death of the latter. One can, therefore, measure 
the rate of adsorption of inactive particles from the survival of bacteria in 
mixtures containing bacteria and irradiated phage in known proportions. 
If, on the average, x particles are adsorbed per bacterium, a fraction e~' of the 



248 



1 1 1 r 




ACTIVE FROM INACTIVATED BACTERIOPHAGE 

T 1 r 



95 



' » I I I I L 




10 20 30 40 10 20 30 40 50 60 70 10 20 30 SECONDS 

Figure 1. — Inactivation of various phages by ultraviolet light. P/Po = proportion of active 
phage particles after irradiation. r = In PoZ-f = average number of lethal hits per phage particle. 
The doses are expressed in seconds of exposure. The deviations for high doses in the curves for 
T2, T4, and T6 are due to reactivation occurring on the assay plates (see text). The broken lines 
represent extrapolations from the logarithmic portions of the curves. 

bacteria should receive no particle and survive. By this means, we could 
establish that for irradiated phages T2, T4, T5, and T6, even for high doses, 
the rate of adsorption is the same as for unirradiated phage. An experiment of 
this type is shown in table 1. Only for very high doses a slight reduction occurs 
in the ability of phage to kill bacteria. This reduction is never such as to require 
important corrections in the analysis presented in later sections. 

With phages T2, T4, T5, and T6 (the "large particle" phages) the plaque 
counts on irradiated samples are not independent of the mode of assay. They 
depend on the concentration of the samples when first mixed with bacteria, 
in a way illustrated in table 2. In these experiments, bacteria were mixed with 
various concentrations of irradiated phage. Before lysis and phage liberation 

Table 1 

Killing of bacteria by irradiated phage 

0.9 ml of a bacterial culture was mixed with 0.1 ml of each of five suspensions of phage T2r 
that had received various doses of radiation. After 10 minutes, samples were diluted and plated 
for viable bacterial count. 





PHAGE 

INPUT, 

PARTICLES 

PER ML 


BACTERIAL 
INPUT, 
CELLS 
PER ML 


DOSE OF RADIATION 


SURVrVING 

BACTERIA 

PER ML 


SURVrVAL 
INPUT 


PHAGE 
ADSORBED 
PER BAC- 
TERIUM 
X 


EXPERI- 
MENT 
NO. 


SECONDS 


HITS 

PER 

PARTICLE 














2.3X108 


0.20 


1.60 








60 


18 


2.4X10^ 


0.21 


1.56 


129 


2X10' 


1.15X109 


70 


21 


2.2X108 


0.19 


1.66 








80 


24 


1.6X108 


0.14 


1.96 








100 


30 


2.7X108 


0.23 


1.47 



249 



96 



S. E. LURIA AND R. DULBECCO 



took place, dilutions were made to bring the total dilution of the irradiated 
sample to a constant value, and an aliquot plated for plaque count. The plaque 
counts represent infected bacteria that liberate active phage. Although the 
total dilution of the irradiated phage on all plates is the same, it is seen that 
the plaque counts are higher when bacteria have first been placed in contact 
with a more concentrated phage lysate. 

This means that bacteria may produce active phage if they pick up the ir- 

Table 2 

Dependence of plaque counts on irradiated phage T6r on the concentration 
of the phage sample that is mixed with bacteria 

A sample of phage T6r containing 1.5X101° particles/ml was irradiated for 20 seconds. The 
bacterial suspension (B) contained 2X10^ cells/ml. Each plate received 0.05 ml of phage dilution 
and 0.2 ml of suspension (B). 











TOTAL DILU- 










DILUTION 


TION FROM 


PLAQUE 
COUNT 

(sum of 

TWO 

plates) 


EXPERI- 
MENT 
NO. 


MIX- 
TURE 
NO. 


PROCEDURE 


OF PHAGE 
WHEN 
FIRST 
MIXED 


THE ORIGINAL 

PHAGE TO THE 

SUSPENSION 

FROM WHICH 








WITH (b) 


SAMPLES ARE 










PLATED 





0.1 ml T6r-^Q.9 ml (B); kept 10 min. 

at 37°C; diluted 1 : 10^, 0.05 cc plated 1 : 10 



1:10^ 



1318 



0.1 ml {T6r 1:10) -^0.9 ml (B); kept 
10 min. at 37°C; diluted 1:10^, 0.05 
ml plated 

0.1 ml {T6r l:103)-»0.9 ml (B); kept 
10 min. at 37°C; diluted 1:10, 0.05 
ml plated 

0.05 ml {T6r 1:10^) plated 



1:10^ 



1:10* 



Less than 

1:105 
(on plate) 



1:10* 



1:10* 



1:10* 



474 



250 



57 



radiated particles from a concentrated phage suspension, but not from a dilute 
one. The immediate explanation is that from a concentrated lysate the bac- 
teria receive some other "factor," which, inside the bacterium, somehow 
reactivates an "inactive" particle and which is not present in dilute lysates. 
An "inactive" particle can be defined as one that has lost the ability to initiate 
production of active phage unless adsorbed by a bacterium together with the 
unknown "factor." 

Reactivation still occurs after storage of irradiated phage for weeks in an 
ice-box. Reactivation gives rise to fully active phage particles. This can be 
proved either by sampling phage from the plaques or by letting the bacteria, 
in which reactivation occurs, lyse in liquid and then testing the lysate for 
active particles. 



250 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 97 

The occurrence of reactivation for certain phages accounts for deviations 
from the logarithmic inactivation rate found for these same phages 
(see figure 1). As the dose of radiation increases beyond a certain point, the 
survival, as determined by plaque count, appears to diminish less rapidly. 
This is due to an unavoidable partial reactivation. In a phage titration, we 
mix approximately 5X10^ bacteria with an amount of phage suspension such 
as to give approximately 100 plaques, and pour the mixture on an agar plate. 
On the one hand, if we are dealing with a fully active sample, the bacteria 
only come in contact with 100 active particles. On the other hand, in 
assaying an irradiated suspension containing, for example, one active 
particle in 10®, we expose the bacteria to 10^ inactive particles plus a corre- 
spondingly large amount of other lysate constituents, besides the residual 
100 active particles. Conditions permitting reactivation, therefore, obtain in 
these plating mixtures, and reactivation disturbs and often completely ob- 
scures the count of the residual active phage. For this reason, the survival of 
fully active phage for high doses must be obtained by extrapolation from the 
logarithmic part of the curve, a necessarily inefficient procedure. If deviations 
in the survival rate for high doses occurred, this extrapolation would not be 
justified. One possible cause of error — screening of some phage particles from 
radiation by components of the lysate itself — was excluded by irradiating 
concentrated phage T6 mixed with phage Tl, and testing for the inactivation 
rate of the latter, which is not disturbed by reactivation phenomena. The 
inactivation rate of Tl remains the same as in the absence of T6 up to doses 
that correspond to 100 hits per particle of phage T6. 

Identification of the reactivating factor 

What is the "factor" present in irradiated stocks of phages T2, T4, T5, or 
T6, which, if acting on bacteria that have adsorbed inactive phage particles, 
allows production of active phage? Since phage stocks are lysates produced 
by lysis of the common host E. coli B, the factor might be either of phage or of 
bacterial origin. 

The factor was identified as being phage itself, inactive or active, in that 
reactivation occurs in bacteria that adsorb either more than one inactive particle of 
a given phage, or one inactive particle of one phage plus some active or inactive 
particles of a related phage. The evidence for this conclusion, which is illustrated 
in part by the data in table 3, can be summarized as follows. 

(a) Addition of an excess of supernatant from a heavy bacterial culture to 
a mixture of dilute irradiated phage plus bacteria gives no increase in plaque 
count. The factor in the lysates is not a normal bacterial secretion. 

(b) Concentrated lysates of phages Tl, T5, or T7 added to a mixture of 
bacteria and dilute irradiated T2 (or T4, or T6) do not cause reactivation. 
No heterologous lysate causes reactivation of T5. The test for cross-reactivation 
is done by plating the mixtures, before lysis, with bacterial indicator strains 
sensitive to the phage whose reactivation is tested, but not to the others. 
Since all phage stocks are lysates of common host cells, the factor is not an 
unspecific bacterial product liberated upon lysis. 



251 



98 S. E. LURIA AND R. DULBECCO 

Table 3 
Reactivation of phage T2 under various conditions 

Phage T2, containing 2X101° particles per ml, was irradiated for 35 seconds. Phages T6 and 
T4 containing 4X10^" particles per ml, were irradiated for 30 seconds. Phages Tl and T5 con- 
tained 3X101" particles per ml. The bacterial suspensions (B) and (B/6) contained 10^ cells 
per ml. 









PLAQUE 




mixture 

NO. 


CONTENTS 


AFTER 10 MINUTES 


COUNT 

(sum op 

TWO 

plates) 


REACTIVA- 
TION 


1 


0.1 ml Phage T2 dil. 1:500 + 1.9 ml (B) 


0.05 ml plated with B 


22 


- 


2 


0.1 ml Phage T2 dil. 1:500 + 1.9 ml (B/6) 


0.05 ml. plated with B/6 


20 


- 


3 


0.1 ml Phage T2 pure +1.9 ml (B) 


dil. 1:500, 

0.05 ml plated with B 


4000 


+ 


4 


0.1 ml Phage T2 pure +1.9 ml (B/6) 


dil. 1:500, 

0.05 ml plated with B/6 


3000 


+ 


5 


0.1 ml Phage T2. dU 1 :500+0.1 ml irrad. 


0.05 ml plated with B/6 


670 


+ 


6 


r<5 + 1.8ml (B) 
0.1 ml Phage T2 dil. 1:500+0.1 m! unirrad. 


0.05 ml plated with B/6 


190 


+ 


7 


T6 + 1. Sm\ (B) 
0.1 ml Phage T2 dil. 1:500+0.1 ml irrad. 


0.05 ml plated with B/4 


661 


+ 


8 


r4 + 1.8ml (B) 
0.1 ml Phage T2 dil. 1 : 500 +0.1 ml irrad. 


0.05 ml plated with B/6 


10 


- 


9 


r<5 + 1.8ml (B/6) 
0.1 ml Phage T2 dil. 1:500+0.1 ml unirrad. 


0.05 ml plated with B/6 


12 


- 


10 


r(5 + 1.8ml (B/6) 
0.1 ml Phage T2 dil. 1:500+0.1 ml irrad. 


0.05 ml plated with B/4 


883 


+ 


11 


14 + 1.8 ml (B/6) 
0.1 ml Phage T2 dil. 1:500+0.1 ml unirrad. 


0.05 ml plated with B/1, 5 


18 


- 


12 


r7+1.8ml (B) 
0.1 ml Phage T2 dil. 1:500+0.1 ml unirrad. 
r5+1.8ml(B) 


0.05 ml plated with B/1, 5 


17 


- 



(c) Concentrated lysates of phages T2 (or T4, or T6), whether fully active 
or irradiated, can reactivate irradiated ptirticles of any other T-even phage. 
These phages are morphologically, serologically, and probably genetically 
related, whereas T5 belongs to a fully separate group (Delbruck 1946). The 
"factor" in the lysates appears to carry the same pattern of relatedness. It is 
not produced by irradiation, since its presence can be proved in unirradiated 
lysates by the technique of cross-reactivation. 

(d) Reactivation is independent of contact between phages prior to infection 
of the host: a mixture of bacteria and phages gives the same amount of re- 
activation independently of how long the phages have been together before 
adding the bacteria. 

(e) Phage T4r purified by fast centrifugation (kindly supplied by Dr. T. 
F. Anderson) shows both self-reactivation as a function of concentration in 
the mixtures, and ability to reactivate phages T2 or T6. The factor, therefore, 
is present in such a purified phage suspension. 

(f) Reactivation of phage T2 by lysates of T6, for e.xample, only occurs in 
presence of bacteria capable of adsorbing both phages. Inactive phage T2 in 
presence of bacteria B/6, by which it is adsorbed, is not reactivated by T6, 
which is not adsorbed, but is reactivated by T4, which is adsorbed. This proves 



252 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 99 

that the "factor" has the same host specificity as the phage in whose lysate 
it is found. 

The last point, particularly, was considered crucial in showing that produc- 
tion of active phage from an inactive particle was actually due to infection of 
the same bacterial cell with other particles, either of the same or of a different 
but genetically related phage, active or inactive. Further confirmation came 
from the experiments discussed below, which showed that in a mixture of 
inactive phage and bacteria the number of bacteria yielding active phage is 
never greater than the number of bacteria that adsorb two or more inactive 
particles, and in some cases actually equals it. The same holds true for re- 
activation of an inactive phage, for instance T2, by a related one, for in- 
stance T4; the number of bacteria liberating active T2, is lower than — -or equal 
to — the number of bacteria receiving at least one particle of each phage. The 
analysis of cross-reactivation between these different wild-type phages will 
not be discussed further in this paper, but will form the subject of a future 
publication. 2 

(g) All bacteria which, after infection with inactive particles, do not liberate 
active phage also fail to lyse. This was proved by mixing bacteria and inactive 
phage under conditions in which some reactivation occurs, plating a sample of 
the mixture for plaque count, and another sample for direct microscopic ob- 
servation of lysis on agar. The results of such experiments proved that the 
fraction of bacteria that are lysed is the same as the fraction of bacteria that 
liberate active phage. The other infected bacteria fail to grow and divide, and 
can be seen still apparently unchanged 24 hours later. 

Reactivation and genetic transfer 

The limitation of cross-reactivation to the T-even phages immediately 
brought out a similarity between this phenomenon and that of genetic transfer 
described by Delbrijck and Bailey (1946). In the latter case, bacteria 
simultaneously infected with the phages T2r^ and T4r — the r character being 
the result of mutation from the wild-type, which can be designated as r+ — liber- 
ate a mixture of particles of the four types, r2f+, T2r, T4r+, and T^,. among 
which the second and third represent new types. These must owe their origin 
to some sort of recombination involving the genetic determinants for the 
alternative r+ and r phenotypes. Evidence for the discrete nature of these 
determinants has since been reported by Hershey and Rotman (1948). 

We assumed then, as a working hypothesis for the analysis of the reacti- 
vation phenomenon, that inactivation by ultraviolet light resulted from 
"lethal mutations" in a number of discrete genetic determinants among in- 
active particles in the same bacterium to reconstitute fully active particles. 
This hypothesis can be formulated quantitatively in terms of measurable 

* Cross-reactivation between T-even phages has the limitation that the individual phages are 
distinguishable only by test of differential properties such as ability to grow on different hosts and 
rate of inactivation by different antisera. Cross-reactivation can only be defined as the produc- 
tion, upon mixed infection, of active particles having the distinctive properties of an inactive 
parent particle. 



253 



100 S. E. LURIA AND R. DULBECCO 

quantities by making a number of simple assumptions. We shall first develop 
this simple theory and then describe the experiments by means of which it was 
tested. 

Theory 

We shall assume that in each particle of a given phage there exist n "units" 
(or "loci") each capable of undergoing a lethal mutation when exposed to 
ultraviolet light. Since one effective hit is sufficient to inactivate a phage par- 
ticle, a lethal mutation can be defined as an effective hit, that is, as an altera- 
tion of one unit which makes the phage particle unable to initiate by itself the 
production of active phage in a bacterium. A particle may undergo more than 
one lethal mutation, and mutations will be distributed at random and independ- 
ently among the different units, the distribution depending only on the 
sensitivity of each unit. 

We shall now make the assumption that the sensitivity of all units is the 
same, and show later that this assumption, if incorrect, only requires a nu- 
merical correction which does not invalidate the applicability of the theory. 

Our next assumption is that active phage cannot be produced in a bacterium 
unless the infecting particle or particles, taken as a group, contain at least one 
copy of each unit in non-lethal form. This assumption is an essential feature of 
the theory, and corresponds to treating each unit as a discrete, material, in- 
dependent hereditary unit endowed with genetic continuity and individuality. 
An inactive unit cannot be replaced by copies of different units. This assump- 
tion implies that production of active units cannot result from the cooperation 
of two or more lethal units, but only from actual reproduction of active units. 

When a population consisting of M phage particles is irradiated with a given 
dose, there will be produced in each particle, on the average, r lethal mutations, 
or a total of MX^ mutations in the whole population. Since M particles contain 
MXn units, each unit will receive on the average Mr/Mn = r/n lethal muta- 
tions. 

A given unit, taken at random, will have a probability e"'''" of not having a 
lethal mutation, and a probability (1 — e"'"''") of having at least one. 

We ask next: what is the probability that each of the n units is present in at 
least one non-lethal copy in a group of k particles that enter a bacterium? 
According to our assumptions, this probability should represent an upper limit 
for the probability that a bacterium produces active phage. 

If a bacterium is infected by k particles, the probability that a given unit 
is lethal in all of them is: (1 — e~'"'")'', and the probability that it is non-lethal 
in at least one of them is: 1 — (1 — e"""'")^. 

The probability that at least one non-lethal copy of each of the n units is 
present in the k particles is the product of the probabilities referred to the 
individual units. Since we have assumed equal sensitivity for all units — that 
is, r/n constant for all units for each value of r — the product will be 

[1 - (1 - e-"-/")^]". (1) 

The expression (1) represents the probability that a bacterium infected by 
k particles receives at least one full non-lethal complement of the n units. 



254 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 101 

In a mixture of phage with bacteria, however, there is a distribution of the 
number of phage particles infecting individual bacteria. With relatively good 
approximation — -see Appendix — this distribution can be considered as a 
Poisson distribution. If x is the average number of particles adsorbed per 
bacterium, the fraction of bacteria with k particles is: x^t~^Jk\ 

The fraction of bacteria receiving k particles which carry a full complement 
of non lethal units is then: 

^— [1 - (1 - e-/")^]". (2) 

Finally, the fraction of the total bacterial population which receives all units 
in a non-lethal form is the sum of the expression (2) for all possible values of )i\ 

Z = X — — tl - (1 - e-/")^]". (3) 

fc=0 k\ 

This expression embodies the following consequences of our hypothesis: 

(a) No full complement of active units can be present in uninfected bacteria 
(^ = 0). 

(b) Of the bacteria with one phage particle (^=1), only those with an active 
particle fulfill the requirement for active phage production (Z = xe~'^e~'') • 

(c) Any bacterium that receives at least one active particle fulfills the 
requirement for active phage production, whether it also receives inactive 
particles or not. For each unit of that particle, 1 — e~''/" = 0; hence, 
(l-e-'-/")* = 0, and [l-(l-e-'-/")^]«= 1. 

(d) For any given value of r>0, Z increases with increasing a;, that is, the 
probability of having a full complement of active units increases as the number 
of particles adsorbed per bacterium increases. 

(e) For any given value of x, Z diminishes with increasing r, that is, the 
probability of having all active units diminishes as the dose of radiation 
increases. 

For the purpose of comparison with data from diflferent experiments, it is 
more convenient to eliminate from the computation those bacteria that receive 
either zero or one phage particle, since they are not expected to contribute 
to reactivation. This is done by using instead of Z the expression 

2 = Z ^^ [1 - (1 - e-ri-rh (4) 

z represents the fraction of bacteria in the total population that have two or 
more phage particles, which together contain a full complement of active units. 
Since the fraction m of bacteria with two or more phage particles ("multiple- 
infected bacteria") is 

m = 1 - {x -\- l)e-^ (5) 

the multiple-infected bacteria receiving a full complement of active units 
represent a fraction 



255 



102 S. E. LURIA AND R. DULBECCO 

2^ [l _ (1 _ Q-rlny^n 

, k=2 kl 

y = z/m = , n -X • (^^ 

1 — (a; + l)e ^ 

The expression y thus obtained is a function of x (average number of phage 
particles adsorbed per bacterium), of r (average number of lethal hits per par- 
ticle), and of n (number of units per particle). 

We shall call w the ratio between the nuniber of bacteria that actually liber- 
ate active phage (plaque count) and the number of multiple-infected bacteria. 
For each mixture of bacteria and irradiated phage, we can determine experi- 
mentally r, X, and the placjue count, and obtain from these the values of m and 
w. We can then compare the experimental values of w with the calculated 
values of y for several different values of n.^ 

The function y = F {r, x, n) was tabulated numerically for a range of values 
of r (between 3 and 50), of x (between 0.05 and 20), and of n (between 10 and 
60). We used for k those ranges of values for which the contributions of the 
corresponding classes were relevant. The corresponding curves were drawn 
iox y = F (r) {x and n constant), and for y = F (x) (r and n constant). 

Before comparing the experimental results with the curves, it is useful to 
discuss briefly what we may expect from the comparison. If reactivation only 
occurs in bacteria with more than one inactive particle, w should never be 
greater than unity. If among the requirements for reactivation there are those 
stated in the assumptions of our theory, the ratio w/y should never be greater 
than unity. Finally, if the requirements stated in our assumptions are necessary 
and sufficient for reactivation, the ratio w/y should be unity, that is, active 
phage should be produced in all those bacteria that receive a full complement 
of the hypothetical units in non-lethal form. Should this obtain, it would then 
be possible to calculate the value of n for each phage from the experimental 
values of w. 

It is important to keep in mind that the assumptions of our theory, up to 
this point, do not contain any implication as to the nature, properties, or 
mechanism of transfer of the postulated units. They only assert that each unjt 
has genetic individuality and can be made lethal by radiation as a result of one 
photochemical reaction, which is of the "all or none" type and independent of 
other reactions of the same type in other units of either the same or other phage 
particles. Production of active phage is conditioned by the presence in one 
bacterium of one active copy of each unit, this copy not being replaceable by 
any number of inactive copies. 

The theory does not imply that all phage particles receive the same number 
of lethal hits, but that the lethal hits are distributed at random among the 
units of all phage particles. 

' In preliminary reports (Luria 1947, 1948) we used the symbol y both for the theoretical and 
experimental probabilities of reactivation. We also gave values of l/y instead of y(or w). The 
present notation, while consistent with the previous one, makes the presentation mere logical. 



256 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 103 

Comparison of theory with experiment 

Quantitative experiments consisted of testing mixtures of bacteria and ir- 
radiated bacteriophage for the number of bacteria that liberate active phage. 
Only phages T2^ T4, and T6 were studied in detail. 

In a typical experiment, such as the one described in detail in table 4, a 
standard culture of bacteria, grown to a titer of either 10** cells per ml or 10^ 
cells per ml, was chilled by immersion in a water bath at 5-6°C. This treatment 
interrupts multiplication without changing the ability to resume immediate 
multiplication and to support normal growth of phage upon return to 37°C. 
In experiments with bacteria grown to a titer of 10^ (latest part of logarithmic 
growth phase), immersion in ice-water is not necessary, since multiplication 
stops almost immediately upon interruption of aeration and transfer to room 
temperature. 

Ten minutes after interrupting multiplication, a sample of the culture is 
diluted and assayed for viable count. If several mixtures of bacteria and phage 
are to be prepared, the culture may have to be used for one or two hours, in 
which case at least one other similar assay is made at the end of the experiment 
to make sure that no proliferation has occurred. At intervals, undiluted samples 
of the culture are placed into test tubes, and a constant volume of irradiated 
(or control) phage variously diluted is added. In this way, we know for each 
mixture the input of bacteria and of phage per ml. For irradiated samples, the 
input of active phage is determined from one or more assays done at very high 
dilution. When this is impossible — for high doses of radiation — the amount of 
active phage is determined by extrapolation from the first part of the inacti- 
vation curve. 

Adsorption is interrupted by heavy dilution after a short time (generally five 
or ten minutes). The amount of adsorption is determined either by determi- 
nation of free phage in the supernatant of a centrifuged sample from a mixture 
containing active phage, assuming similar adsorption in all other mixtures 
(see table 1), or by determining the bacterial survival in samples from various 
mixtures. With the T-even phages, 80 to 95 percent of the phage is adsorbed 
in ten minutes. The value for the multiplicity of infection, x, for each mixture 
is used in calculating the fraction m of bacteria with two or more phage par- 
ticles: m=l — (x-\-l)e~'. 

Before lysis begins, a suitably diluted sample of the mixture is plated with an 
excess of sensitive bacteria for plaque count. The plaque count — corrected, 
when necessary, for the active phage by subtracting the value corresponding 
to the latter — is divided by m to obtain the experimental value w for that 
mixture (see table 4). 

A large number of experiments, yielding a total of over 1000 values of w, 
for various doses of radiation and for different multiplicities, were done with 
the T-even phages. Phage T5 was only partially investigated, because of diffi- 
culties in obtaining reproducible results in view of the low and irregular adsorp- 
tion rate for this phage. 

The data for phages T-even include those for several of their r mutants, 



257 



104 



S. E. LURIA AND R. DULBECCO 



to. 






^ e* 



O ui 



■ CN 
^ II 
II ^ 

2X 

^ II 



>'E 

3 . 

2X 
X^ 

c°2^ II 

O Jj CM (U .ti 

So S ^ I 

u O . "2 D. 

— -S 2^00 u, 

rt _. J2 ■* <u 

m ii _ 4J 60 

CO n! _ ^ 
. l^ (U t- . 

II ^ -p -2 
^•«-2X 

^ ■^ T3 12 
<L) U . . ^ T3 

s 00 .rt r . 
■?'-'_ u o 
t-i o aj <« <« 



O '^l D. Uh 1-. 

S (U u CJ <u 

^ fcC*j M bC 

.—4 c4 C cd c^ 



5, '-' 

W Q < 

& U W 

H w H 

J £ o 

^ z ;S 

S S ea 



H fc! & z 
3 u g o « 



Q 






O X 






d2 **§ S^ 
pi, ^ z o t> a 



X « 6 
3 R z 



Q^r»5 t--^.^T-H 06rt<OO lOCNO 
OO OOO OOOO OOO 



O OOO OOO 

22 X2X XXX2 XXX 

-■ oo ro r^ N/ ~" " 



XX "IX 

0\«Ni r^ .rt cs 00 '^ ■* ■<* 



-* r<3 •^ 
1^3 CN lO 



OO 



OO 



2x 2XX 22XX 22x 
x*^ X"^.^- XX"1^ xx^. 

,— (lO »-ilC'— I — ' — '"^ — ' — 1— - — 



I \n ■^ 



(N O 


lo o 

CN O O) 


m o 

LO CN O CN 


20.5 

8.2 

0.20 


00 CN 


00 CN O 


O 00 CN O 



OO OOO OOOO OOO 

XX XXX XXXX XXX 

ir> in ^n in u-> If) m *f^ 

cnO cnoO ocnoo ocno 

00 <^ 00 cn <>1 cn 00 cn cn cn 00 cn 



OO OOO OOOO OOO 

XX XXX XXXX XXX 

00^^ OOCNf^ CNOOCNCN CNOOCN 

OCJOI OOCSCN CSOOCSCN CNCOCN 



fi o o o o o 



oo 
XX 


2. S 
x2x 

LO y CN 


r- t- <£» 

OOO 

XXX2 
(X ^ t^ y^ 


r- r-. CO 

OOO 

XXX 

Tj> ro -^ 


XX 


t^ rt CN 


00 '^ ■* -^ 


u-j CN lO 


<M -H 



.pq wmM 



"lo 



SEE 

Os Os <^ 

d d o 
+ + + 



'e'e'e 



"c dd d 

o\ ^~., — ,, — , 
• lo o 

o . O o 



P3P9 



- — -0^ 0^ 

"5 d d 

f + + 

o"^8 



^3 -a -T3 T3 "O 



So " 

oo 

CMCN 



cj a> ^ 



b u li H 



bo bc 

-^# 
PnFt 



<u 1) l> 

be bc bO 
rt cd c^ 

-3-^:5 
fin PL, Oh 



(U (U (U qj 

bJ3 bC bC OC 
d d d d 

Oh C^ Ph Ph 



1) 4> <U 

bC bC W) 

Ph Ph Ph 



SB BEE 



do OOO 



E E E E 



E E E 
odd 



vO r^ oo On 



O '-I CN 



0^ 

d 

+ 



O O CJ O tj o u 



c 


c 


r> 


ni 


U 


ai 




C 


bC 

Ph 


o. 

3 








C 


E 


<u 


■^ 


<u 


o 


tn 



X=^ 

<M O 

(-j II 

II '^ 

Or) 
ro 5 O 

►S^ ^ 

^ aj C 



X 
m 

00 

g'ii 

O <U 
O II rt 

X.2.3 



„ ^^ 
u, ►^ o 



2b'5 

Xyt 

O '^ 

II <^.2 
.i llx 

<u S <" 

V C J- 

-L- '" a 
+ a? 

+-» 4-1 '2 

rt 3 C 

-TH O <U 



258 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 105 

Table 5 

Probability of reactivation for various phages as a function of 
dose and of vudtiplicity of infection 

Values of the ratio iv between bacteria that yield active phage and multiple-infected bacteria. 



EXPERI- 
MENT 
NO. 


PHAGE 


MULTI- 
PLICITY 
X 


AVERAGE NUMBER OF HITS PER PARTICLE 

r 2.6 4.8 6.0 9.5 10.0 12.5 


15.8 






0.2 








0.14 




0.025 


0.0034 






0.4 












0.028 


0.0055 






1 












0.055 


0.01 






2 






0.36 


0.18 




0.085 




155 


T2r 


4 












0.18 


0.08 






8 






1.0 


0.83 




0.48 


0.26 






20 












0.98 


0.6 






0.1 




0.56 


0.42 




0.1 










0.2 


(1.0) 


(0.83) 


0.40 




0.1 










1 


(0.9) 


(0.59) 


0.44 




0.17 










2 






0.5 




0.23 






156 


T2r 


4 

8 

20 






0.67 
1.05 
1.0 




0.43 
0.74 
1.0 










r 3.6 


6.1 


7.9 


9.6 


11.7 


12.8 


15.5 




0.2 








0.065 












0.4 




0.15 




0.075 




0.03 








0.8 




0.21 




0.1 




0.035 




162 


T6r 


2 
4 

8 




0.27 
0.59 
0.83 




0.18 
0.26 
0.67 




0.05 
0.18 
0.37 








0.35 


(1) 




0.27 




0.07 




0.016 






0.7 


(1) 




0.37 




0.08 




0.016 


70 


T6 


1.4 

2.8 
5.6 






0,37 
0.37 
0.48 




0.12 
0.14 
0.21 




0.04 
0.06 
0.08 




r 


5 






10 








0.35 




0.4 






0.027 






160 


T4 


0.7 
1.4 
2.8 

7 
14 




0.29 

0.28 

0.42 

0.8 

1.0 






0.042 

0.06 

0.13 

0.38 

0.77 







The values in parentheses are from separate experiments. 



259 



106 S. E. LURIA AND R. DULBECCO 

which were found to have the same probability of reactivation as the respective 
wild types. The results cover ranges of values of r from 2.5 to over 30, and of 
X from 0.02 to 20. 

The individual values of w, and those of the variables, r and x, are obtained 
from the following actual measurements: 1) titer of phage; 2) total number of 
bacteria; 3) survival of phage; 4) survival of bacteria and/or assay of free 
phage; 5) plaque count from the mixture. Each of these measurements involves 
an error of estimation due to dilution and sampling errors. Several of these 
determinations, however, are the same within each experiment. The results 
from individual experiments are, therefore, more consistent than those from 
different experiments, as shown in table 5. 

The only graphic representation that could show all values of w for each 
phage and allow of comparison with the calculated values of y would be a 
tri-dimensional plot of w as a function of r and of x. As second best choice, 
we plotted the values of w as a function of r for several values of x taken as 
constant, and as a function of x for several values of r taken as constant. 
Individual values of w fluctuate rather widely, but the data as a whole make 
it possible to draw curves, which represent averages and which can be consid- 
ered as the curves for 2i> as a function of r and of x. A number of such plots 
using all the experimental points for the corresponding values of the variables, 
are presented in figures 2,3, and 4 (multiplicity of infection as variable) and 
figures 5, 6, and 7 (dose of radiation as variable). 

The trend of these plots is similar to that of the theoretical curves for y^ 
and it is possible to find for each phage a constant value of n (number of units) 
such that the corresponding values of y become very similar to those of w for 
low values of x and for any value of r. That is, it is possible for each phage to 
determine a constant number of units for which the experimental probability 
of reactivation equals the theoretical one for any dose of radiation provided 
the multiplicity of infection is low. The corresponding theoretical curves for 
y have been drawn in the plots of the values of w. For phage T2^ the best fit is 
for w = 25; for T4, n=\S\ for T6, w = 30 (see especially figures 5, 6, and 7). 

The main feature emerging from the curves in figures 2, 3, and 4 is that the 
values of w tend to unity for increasing multiplicity of infection. In several 
cases, the number of cells that liberate phage actually reaches the number of 
multiple-infected cells, but in no case does it go beyond it, proving that reacti- 
vation does not occur in single-infected cells. 

The curves in figures 5, 6, 7, for w as a function of r, are of the multiple-hit 
type, indicating that suppression of phage production depends on damage in a 
number of elements. The values of w tend to unity for low doses, again showing 
that reactivation potentially can take place in every multiple-infected cell. 

Comparison in figures 2-4 with the curves for y, chosen to fit the experi- 
mental curves for low values of x, shows that the general similarity is limited 
by a systematic deviation. As the multiplicity increases, both w and y tend 
asymptotically to unity, but w increases more slowly. This means that, as the 
number of phage particles per bacterium increases, the probability of reacti- 



260 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 



107 




Figure 2. — The probability w of reactivation for irradiated phages T2 and T2r as a function 
of the multiplicity of infection x, for several doses of radiation. Abscissae: values of x. Ordinates: 
values of w. 

# values of w for r= 6 hits per particle. 

+ values of w for r= 10 hits per particle. 

O values of w for r = 20 hits per particle. 
Solid lines: theoretical curves for y as a function of x for n = 25, and for the values of r given above. 



261 



108 

w 



S. E. LURIA AND R. DULBECCO 



10" 



10 



-2 



10" 



10- V. 







' 


■ 1 




• 




/ 




O 

o 

o 


o 


e 

O J 

e X 
e / 








o j 




T4 




/o 









Figure 3. — Same as 



jure 2, for phages T4 and T4r. The theoretical curves 
for y correspond to n = 15. 



vation, while steadily increasing, does not keep pace with the theoretical 
function y. The cooperation within groups consisting of more than two par- 
ticles is not as successful in bringing about reactivation as required by the 
simple theory, whereas pairs of particles apparently collaborate with an 
efficiency of one hundred percent. 

The deviation for higher multiplicities is reflected in the curves of figures 5-7, 
The curves for w as a function of the dose are very close to the theoretical 



262 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 109 

curves for low multiplicities, up to :*; = 0.5; for higher multiplicities they fall 
below the corresponding curves for y calculated for the same number of units. 
We can actually find, for each value of x, a curve for y — for the same n but 
corresponding to a lower x — which fits the experimental curve. The correspond- 
ing theoretical curves are drawn in figures 5-7. This indicates that groups of 
more than two particles collaborate in reactivation as if they consisted of a 
lower but definite number of particles. 

It is interesting to notice that the systematic deviation from theory for 



W I 




Fgure 4.— Same as figure 2, for phages T6 and T6r. The theoretical curves for 
y correspond to n = 30. 



263 



S. E. LURIA AND R. DULBECCO 




Figure 5. — The probability w of reactivation for phages T2 and TZr as a function of the dose 
of irradiation r (in hits per particle) for several multiplicities. Abscissae: values of r. Ordinates: 
values of w. 

# values of w for a: =0.1-0.2. 

O values of w for a; =0.8-1. 5. 

+ values of w for x= 2.5-4.0. 

Broken line: theoretical curve for y as a function of r for n = 25, a:=0.15. Solid line: theoretical 
curve for y as a function of r for n = 25, a;=0.6. Broken and dotted line: theoretical curve for y 
as a function of r f or « = 25, a: = 1 .3. 



264 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 



111 



W I 




Figure 6. — Same as figure 5, for phages T4 and T4r. 

# values of w for a;=0.1-0.2. 
O values of ui for a; =0.8-1. 5. 
+ values of w for x= 4-7. 

Broken line: curve of y for « = 15, a: =0.1 5. Solid line: curve of y for m = 15, .t= 1.1. Broken and 
dotted line: curve of y for m = 15, a;=4.0. 



265 



S. E. LURIA AND R. DULBECCO 




5 10 15 

Figure 7. — Same as figure 5, for phages T6 and T6r. 

# values of w; for :c=0.1-0.2. 
O values of w for a; =0.8-1. 5. 
+ values of z« for a; =2. 5-4.0. 

Broken line: curve of y for « = 30, a;=0.15. Solid line: curve of y for w = 30, a;=0.5. Broken and 
dotted line: curve of y for m = 30, a;=0.9. 



266 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 113 

increasing multiplicities is less evident for phages T2 than for T6, and still 
less for T4, diminishing in the same order as the calculated number of units. 

To summarize, the results show that the probability of production of active 
phage from inactive particles depends on the dose of radiation and on the 
multiplicity of infection in a way similar to the one predicted by the simple 
theory, which assumes lethal mutations in discrete transferable units of equal 
radiation sensitivity and a hundred percent efficient recombination of active 
units to reconstitute active particles. All deviations can be accounted for by a 
limitation in the efficiency of recombination when the active units derive from 
more than two inactive particles. 

Several explanations may be offered for this limitation; some of them have 
been tested experimentally. Lysis from without — ^failure to liberate phage due 
to excessive multiplicity of infection (Delbruck 1940) — was found not to 
take place for multiplicities of the order of those for which the deviations from 
theory occur. Limitations in the number of particles of a given phage that can 
participate in phage growth were looked for and found (see Dulbecco 1949a), 
but their magnitude cannot account for the differences between w and y. 

The process of reactivation must involve complex mechanisms of transfer 
of genetic material among phage particles. Whatever these mechanisms, it 
is reasonable to expect that they will work less efficiently as the number of 
phage particles increases. Limitations may conceivably be caused by steric 
reasons — shape of the particles, position in the bacterium — ^or by physiological 
reasons — limited number of units of some catalyst, competition for substrates. 

Estimation of the number of units 

We have compared our results with the theoretical curves for y calculated 
for different values of n, the unknown number of transferable units per particle. 
The curves that best fit the results for phages TI and T2r are those for « = 25; 
for T6 and T6r, « = 30; for T4 and T4r, w=15. For phage T5, no accurate 
estimate of n was obtained, but n appears to be lower than for T4. 

If the interpretation of the results based on the simple theory is justified, 
we must consider the values thus obtained for n as minimum estimates of the 
number of radiation-sensitive, transferable units per phage particle. The 
estimates are minima because of the assumption of equal sensitivity of all 
units. Should there be units more sensitive than others, they would be hit more 
often, and in order to obtain the correct probability of reactivation we should 
assume more units of the less sensitive type. For example, if one unit were 
twice as sensitive as the average of the others, one locus of the average sensi- 
tivity should be added to our estimate in order to distribute the probability 
of inactivation over all units in such a way that the reactivation probability 
remains the same. 

In our preliminary report (Luria 1947) we calculated w in a different man- 
ner, by assuming that for low multiplicities and low doses, where the probabil- 
ity of reactivation appeared to be approximately constant as a function of », we 
could consider all multiple-infected bacteria as double-infected. This corre- 



267 



114 S. E. LURIA AND R. DULBECCO 

sponded to putting ^ = 2 in formula (6) and to comparing the experimental val- 
ues of w with the values of y for x = 0. Upon closer analysis, this method proved 
incorrect, because the contribution of higher multiple infection cannot be 
neglected, even for low multiplicities. Analysis of more data showed that the 
probability of reactivation is in fact not constant for low values of x, but ap- 
pears to be so for low doses, because the differences are small and of the order 
of the experimental errors. The use of the wrong approximation made our 
previous estimates of n too high. 

Yield of active phage from bacteria in which reactivation occurs 

The yield of active phage following reactivation was studied systematically 
for phages T2r and T4. After infection, bacteria were diluted and allowed to 
lyse in liquid, as in a typical "one-step growth" experiment. The latent period 
before lysis is somewhat longer than for active phage (about 26 minutes instead 
of 21 for T2, 30 minutes instead of 25 for T4), and the rise in phage titer upon 
liberation somewhat slower. All yields were calculated from plaque counts after 
the titer had reached a steady level. The results, shown in table 6, indicate 
that the yields are generally somewhat lower than those from bacteria infected 
with active phage particles. No clear relation of yield to dose of radiation or 
to probability of reactivation was detected. For T2r irradiated with high doses, 
there is a certain tendency toward higher yields for higher multiplicities. 

Mixed infection with active and inactive phage 

Transfer of genetic material involved in reactivation must occur between 
active and inactive phage particles, since, as we saw before, an active particle 
of a T-even phage can reactivate an inactive particle of another T-even phage. 
If transfer occurred by reciprocal exchanges of genetic material, we should 
expect that upon mixed infection with active and inactive particles of the 
same phage some of the active particles would receive inactive units and, 
therefore, be inactivated. This possibility was tested for phages T2 and T4 by 
experiments of the following type. 

Bacteria are added to mixtures containing various proportions of active 
phage and of phage of the same strain irradiated with different doses. For 
each mixture, the average numbers of active and of inactive particles adsorbed 
per bacterium are calculated and, hence, the number of bacteria receiving both 
active and inactive phage. A plaque count before lysis gives the number of 
bacteria that liberate phage, while a plaque count after lysis gives the yield of 
phage per bacterium. In this manner, we can determine whether inactive phage 
suppresses production of active phage from bacteria that also adsorb an active 
particle, or possibly affects the yield. 

The results of these tests can be listed as follows: 

(a) a bacterium receiving an active particle plus one inactive particle of 
the same phage — no matter how many hits the latter has received — never fails 
to liberate active phage; 

(b) part of the bacteria that adsorb one active particle plus several inactive 



268 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 115 

ones (the latter carrying enough lethal hits, so that reactivation does not 
occur among them) fail to liberate active phage. This suppression of active 
phage production is evident for multiplicities 5 or higher of inactive phage T4, 
and for multiplicities 8 or higher of inactive phage TZ. The suppression is in- 
dependent of the time allowed for adsorption of the phages. It may in part be 

Table 6 

The yield of active phage from bacteria in wliich teactivation takes place 



MULTIPLICITY 
OF INFECTION 



12 18 



0.2-0.25 114 51 122 

0.5-0.6 50 42 22 

2.9 89 100 29 



T2r 


7.7 




59 






60 




9 




100 


70 


94 






10-11 
0.65 




73 


120 


59 


130 




HITS 





7.5 


15 


18 






260 


120 


87 


200 




1.1 




200 


87 


162 




T4 


1.6 

2.7 

7-8 

14-15 




275 
212 




140 
197 

135 
250 


140 

200 

85 



accounted for by the limitation phenomenon described by the junior author 
(DuLBECCO 1949a) ; 

(c) for those bacteria that liberate active phage, the yield per bacterium is 
not affected by the presence of inactive particles, but remains the same as in 
controls without the inactive phage; 

(d) when bacteria are infected with several heavily irradiated particles 
that do not give reactivation, and a few minutes later with one active particle, 
suppression of phage production occurs in a proportion of bacteria that 
increases with the interval between infections. Suppression is evident with an 
interval of 2.5 to 4 minutes and practically complete after 10 minutes. These 
time intervals between infections are, of course, averages, since infection may 
occur earlier or later for individual bacteria in the same mixture. The yield 
from those bacteria that liberate phage still remains normal. 

The suppression of active phage reproduction by inactive phage in excess 
indicates the existence of some type of "mutual exclusion" between particles 
of the same phage. Such exclusion is also indicated by the experiments of 
DuLBECCO (1949a). 

More detailed analysis of the interaction between active and inactive 
phage, using genetic markers, will be reported in future papers. Our results 



269 



116 



S. E. LURIA AND R. DULBECCO 



discussed above are in agreement with earlier observations (Lukia and 
Delbruck 1942) on interference by a large excess of irradiated phage T2 with 
the growth of active phage T2 when the inactive phage was mixed with 
bacteria one minute and a half before the active one. 

Cross-reactivation between phage particles dijfering by one character 

Only one group of experiments will be discussed here, because of its bearing 
on the analysis of the data presented in this article. Bacteria were infected with 

Table 7 

Cross-reactivation between one inactive particle of phage T2 and 
one inactive particle of phage T2r 

Compare column (3) with column (1) 







MULTIPLIC- 


MULTIPLIC- 


EXPERI- 


DOSE OF 










ITY OF 


ITY OF 


MENT 


RADIATION, 










INFECTION 


INFECTION 


NO, 


HITS 










FOR T2 


FOR T2r 



(1) 

FRACTION OF 
BACTERIA RE- 
CEIVING one 
INACTIVE PAR- 
TICLE T2 AND 
one INACTIVE 
PARTICLE T2r 

AMONG THE 
BACTERIA THAT 
LIBERATE AC- 
TIVE PHAGE, 
CALCULATED 



(2)* 

OTHER BAC- 
TERIA THAT 
COULD GIVE 

MOTTLED 

PLAQUES (as 

FRACTION OF 

THE BACTERIA 

THAT LIBERATE 

ACTIVE phage), 

CALCULATED 



(3) 



FRACTION 

OF MOTTLED 

PLAQUES, 

FOUND 



lA 


2.9 


0.055 


0.055 


0.19 


0.045 


0.14 


2A 


4.6 


0.055 


0.049 


0.335 


0.053 


0.16 


3A 


6.2 


0.055 


0.055 


0.42 


0.074 


0.155 


4A 


3.4 


0.053 


0.053 


0.24 


0.046 


0.14 


5A 


5.0 


0.053 


0.053 


0.355 


0.068 


0.20 


7A 


3.9 


0.05 


0.05 


0.28 


0.046 


0.14 


8A 


6.5 


0.05 


0.05 


0.19 


0.087 


0.21 



* The values in this column include all bacteria with two or more particles of one type and 
one or more of the other type, plus all bacteria with an active particle of one type and an in- 
active particle of the other. The values are upper limits, since only a fraction of these bacteria 
will actually give mottled plaques. 

phages T2 and T2r, both irradiated (r = 5 or 6). Low multiplicities were used, 
so that a large proportion of the infected bacteria only received one inactive 
particle, and, of those that received two, a great proportion received one 
particle of each type. The infected bacteria were plated before lysis, and the 
plaques examined for the proportion of "mottled plaques," that is, of plaques 
containing both T2 and T2r active phages. Such plaques can only arise from 
bacteria infected with both phages. It is seen from the data shown in table 7 
that more than half the bacteria infected with one inactive particle of each of 
the two phages actually liberate a mixture of active particles of both types. 
This proves that recombination cannot result from reciprocal exchanges of 



270 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 117 

units between the infecting particles bringing together into one particle all the 
active units before multiplication begins. Evidently, this particle should be 
either T2 or T2r, and could not give rise to active particles of both types. This 
conclusion will be analyzed further in the discussion. Quantitative analysis of 
the number and contents of mixed yields from mixed infection with T2 and 
T2r shall be the subject of future publications. 

Phages inactivated by X-rays or nitrogen mustard 

In a preliminary article (Luria 1947) it was stated that no reactivation had 
been detected for T-even phages inactivated by hard X-rays, using the same 
technique employed for ultraviolet. The same was found in our laboratory by 
Miss M. E. Willis for phages T2 and T6 inactivated by a nitrogen mustard 
(methyl bis (/3-chloroethyl) amine hydrochloride). 

Experiments by Mr. J. Watson, still in progress in our laboratory, have 
recently shown, however, that phage T2 inactivated by hard X-rays can take 
part in reactivation, but this reactivation occurs with such a low probability 
that special techniques are required for its detection. In part, the low probabil- 
ity of reactivation of X-ray inactivated phage is due to reduced rate of adsorp- 
tion. This work will be reported by Mr. Watson in a future publication. 

It seems possible that phages TJ and T7, for which no reactivation was 
detected after ultraviolet inactivation, may also be found by similar techniques 
to give some reactivation. It is clear that the probability of reactivation will be 
low if the number of transferable units is small, or if each lethal mutation in- 
volves several units. Its detection will be difficult whenever the number of bac- 
teria in which reactivation occurs is small in comparison with the number of 
bacteria that receive residual active phage.^ 

discussion 

The experiments described above have given results consistent with the 
hypothesis that inactivation of several and possibly all bacteriophages by ul- 
traviolet light is to be attributed to lethal mutations in discrete units of genetic 
material. A genetic basis for inactivation of viruses and bacteria by radiation 
has often been postulated either on statistical grounds or by analogy (see Rahn 
1929; Lea 1947). Our results bring new support to this view, and suggest that 
most, if not all, the inactivating effect of ultraviolet light (2537 A) on certain 
phages is due to the production of localized lethal mutations. 

The hypothesis of inactivation by lethal mutations and reactivation by 
transfer of genetic material following multiple infection, as developed in this 
paper, has been useful in suggesting a quantitative analysis of the reactivation 
phenomena and has led to fairly accurate predictions of the experimental re- 
sults. The following discussion assumes the correctness of this working hy- 
pothesis. 

* While this paper was in press, the senior author found that some reactivation by multiple 
infection takes place with phage Tl inactivated by ultraviolet light. For equal multiplicity of in- 
fection and equal number of hits, the frequency of reactivation is much lower with Tl than with 
any of the T-even phages and with T5. Assuming that the type of analysis presented in this paper 
applies to the results with Tl, a value of n smaller than 5 would be obtained. 



271 



118 S. E. LURIA AND R. DULBECCO 

According to our analysis, active phage is reconstituted from inactive by re- 
incorporation of active units derived, directly or indirectly, from the inactive 
particles in a kind of hybridization. As in hybridization, the possibility of 
recombinations between particles of different wild-type phages suggests the 
existence of common genetic determinants and the absence of complete incom- 
patibility. It is possible that various interference phenomena among different 
phages may result from such an incompatibility. 

Reassembly of material from inactive particles into active ones is a remark- 
ably efficient process. Genetic material from several inactive particles may be 
brought together, although the relative efficiency of cooperation diminishes as 
the number of particles involved in this pluriparental reproduction increases. 

We have given estimates for the minimum number of transferable units per 
particle for several phages. It is interesting to notice that phage T4, more resist- 
ant than the related phages T2 and T6 to ultraviolet light (figure 1), appears 
to have fewer units. This may indicate absence of a portion of genetic material 
present in the other T-even phages. 

Lea and Salaman (1946), analyzing the dependence of the rate of inacti- 
vation of phages by X-rays as a function of the density of ionization, and 
assuming that the radiosensitive material consisted of spherical units, arrived 
at the conclusion that a large phage contained 14 such units, whereas a small 
phage contained one only. Although the hypotheses involved were probably 
oversimplifications, the conclusion receives qualitative support from our re- 
sults. 

We must consider next the possible mechanisms of genetic transfer. We may 
divide the mechanisms into two groups, those in which the reproducing element 
is supposed to be at all times the phage particle as a whole, and those in which 
the reproducing elements are assumed to be component parts of the particle. 

The simplest hypothesis of the first group would be that reactivation 
results from pairing (or grouping) or the initial infecting particles, followed 
by reciprocal exchanges such as occur in chromosomal crossing-over; if these 
exchanges lead to formation of an active particle, the latter proceeds to multi- 
ply. This simple hypothesis can easily be disproved. The high efficiency of 
reactivation would require very large numbers of successive reciprocal ex- 
changes to bring together all active genetic material before multiplication takes 
place. Mixed infection with one active and one inactive particle should also lead 
to the occasional loss of active phage, since we know that genetic recombina- 
tions occur between active and inactive particles. Finally, the fact that in- 
fection with one particle each of inactive T2 and inactive T2r yields a mixture 
of active T2 and active T2r disproves this hypothesis, since any number of 
reciprocal exchanges between the original particles before multiplication could 
never lead to formation of active particles of both types. 

Another interpretation based on reciprocal exchanges would be that inactive 
particles reproduce, and that exchanges occur at various stages of the repro- 
duction among the original particles or their inactive offspring. These ex- 
changes should be numerous enough to make the probability of incorpo- 
ration of all active units into an active particle close to unity, and an ac- 



272 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 119 

tive particle, once formed, should be favored in multiplication. Without 
completely disproving it, our results make this explanation very unlikely. 
The fact that the probability of reactivation increases greatly by increas- 
ing (for example, from 8 to 20) the multiplicity of infection with heavily 
irradiated particles would require a very high number of exchanges. To obtain 
high yields of active phage in these cases we should assume, moreover, that 
the exchanges take place early, and that the active particles, once formed, 
multiply with little or no interference from the inactive particles in large excess. 
This seems contradicted by the fact already mentioned that active particles 
can actually undergo interchanges with inactive ones. Altogether, there is 
strong evidence that inactive units have less chance than active units of enter- 
ing the final particles, even when the active units derive from inactive particles. 

In search for a mechanism that could selectively bring together the active 
units, the senior author (Luria 1947) suggested the hypothesis of independent 
reproduction of individual units to form a "gene pool," from which the new 
active particles could be derived. Inactive units were considered to be those 
that cannot reproduce and that have, therefore, little chance of incorporation 
into the final particles. The tendency to reduction in yield may be due to oc- 
casional incorporation of some of the original inactive units. No hypothesis 
is made as to how the units reproduce or reassemble. The last step is the most 
difficult to visualize, and we incline to the belief that the original particles 
may play a role in it, possibly by supplying a framework for reassembly. This 
is suggested by the limited efficiency of collaboration among large groups of 
particles, which indicates a certain tendency of the units to remain together 
with their original companions. 

One may ask whether, inside bacteria in which reactivation does not take 
place, the active units present in the infecting particles reproduce or not. 
Cohen (1948) states that no desoxyribose nucleic acid is synthesized in bac- 
teria infected with particles of T2 exposed to doses of ultraviolet light much 
higher than those employed in our study. Preliminary cytological evidence, 
collected with the collaboration of Dr. C. F. Robinow in our laboratory, 
indicates that in infected bacteria, in which reactivation does not take place, 
there is no accumulation of stainable material supposedly representing des- 
oxyribose nucleotides. If this evidence is confirmed and found to apply to the 
conditions of our experiments, it might then suggest, either that there is no 
reproduction of active units when they are not all present (which might al- 
together invalidate the hypothesis of independent reproduction), or that at 
least part of the reproduction of the active units may take place without 
increase in desoxyribose nucleotides. 

The hypothesis of a "gene pool," although by no means the only possible 
one,^ fits all results of reactivation. Its validity may soon be amenable to 

^ Another possibility, suggested by Dr. A. H. Sturtevant, would be a process of zipperwise 
replication of the various units of a phage particle. When in this process an inactive unit was 
reached, replication could only continue if the partial replica came in contact with another phage 
particle in which that unit was active. The process would then continue by addition of replicas 
of the active units of the second phage particle. If repeated several times, such a mechanism 
would provide for selective recombination of all active units. 



273 



120 S. E. LURIA AND R. DULBECCO 

critical test. In postulating a phase in phage growth in which particles, as we 
know them in the extracellular phase, are not present, the hypothesis accounts 
for the repeated failures to obtain active phage by premature artificial dis- 
ruption of infected bacteria. It agrees with the observation made by Foster 
(1948) in our laboratory that in the presence of proflavine the reactions leading 
to production of active phage proceed normally for a part of the latent period, 
but, upon lysis, no active particle is liberated. According to Foster (1948) 
active phage particles are present in the infected bacterium after the proflavine 
sensitive stage is passed — beginning 12 to 14 minutes after infection for phages 
T2. ox T6. k similar conclusion was reached by A. H. Doermann on the basis 
of experiments on the effect of other inhibitors on the growth of T3 and T4r 
(Doermann 1948). That viruses multiplying inside the host cell may not have 
the same organization as in the extracellular form has been suggested before 
for certain animal viruses (see Bland and Robinow 1939). 

Our theory does not assume any degree of linkage among units, although 
linkage may be compatible with the theory. Each group of strongly linked 
units would behave as one unit, possibly as a particularly sensitive one. Weakly 
linked units would probably reduce the probability of reactivation for high 
doses, when lethal units present in one group would hinder the utilization of 
the linked active units for reactivation. 

In their work on the r and h mutants of phage T2 H, Hershey and Rotman 
(1948, 1949) found evidence for a series of determinants exhibiting various 
degrees of linkage, from those apparently unlinked (high frequency of inde- 
pendent transfer, no correlation between the frequencies of complementary re- 
combinant types in the yield) to others rather strongly linked. For the latter 
ones, the authors considered that their results suggested the possibility of re- 
ciprocal exchanges. 

As a working hypothesis we may assume, together with Hershey and Rot- 
man (1949) that unlinked determinants may be located in different reactivation 
units, possibly transferred by a gene pool mechanism, while linked determinants 
may be located in the same reactivation unit. If reciprocal exchanges were 
found to occur, then homologous units or groups of units should be supposed to 
pair or group together at some stage in the growth process. Techniques recently 
developed in our laboratory should soon permit a study of the inactivation of 
individual genetic determinants and a solution of some of these problems. 
It appears, therefore, advisable to refrain from further discussion at the present 
time. 

The formation of active phage from inactive by transfer of discrete units 
requires some revision of the interpretation of experiments on irradiation of 
phage inside infected bacteria with ultraviolet light (Luria and Latarjet 
1947). In case of multiple infection, the survival curves for phage-producing 
ability immediately after infection indicated suppression by damage of a num- 
ber of centers, with the sensitivity of individual centers lower than that of 
extracellular phage particles. It now seems clear that what was measured was 
the rate of inactivation of individual units rather than of whole particles. 
The curves for suppression of the phage-producing ability of multiple-infected 



274 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 121 

bacteria immediately after infection are similar to the curves for the probabil- 
ity of reactivation for a comparable group of irradiated particles. For example, 
the suppression curve given by Luria and Latarjet (1947) for bacteria 
infected by five particles of phage T2 is very similar to the curve for the prob- 
ability of reactivation w as a function of r for x = 5. It is clear that in interpret- 
ing experiments on inactivation of intracellular phage it will be necessary to 
take into account the occurrence of genetic transfers.* 

SUMMARY 

Coli-bacteriophages TZ, T4, T5, and T6 inactivated by ultraviolet light 
are still adsorbed by sensitive bacteria. Bacteria infected by only one inactive 
phage particle are not lysed and do not yield active phage. Infection of bacteria 
with more than one inactive particle leads to lysis and production of active 
phage in a fraction of the bacteiia. This fraction diminishes with increasing 
doses of radiation and increases with increasing numbers of particles adsorbed 
per bacterium. The assumption is made that inactivation is due to lethal 
mutations in a number of genetic "units" of the phage particle, and that pro- 
duction of active phage from inactive is due to recombination of non-lethal 
units to form active particles. The values of the probability of active phage 
production calculated from these assumptions agree with the experimental 
results with certain limitations. In order to explain the very high frequency 
of recombination, the hypothesis is proposed that phage growth occurs by 
independent reproduction of each unit followed by reassembly of the units into 
complete phage particles. The minimum number of units per particle is esti- 
mated for various phages. 

LITERATURE CITED 

Bland, J. O. W., and C. F. Robinow, 1939 The inclusion bodies of vaccinia and their relation- 
ship to the elementary bodies studied in cultures of the rabbit's cornea. J. Path. Bact. 48: 
381-403 

Cohen, S. S., 1948 The synthesis of bacterial viruses. I. The synthesis of nucleic acid and protein 
in Escherichia coli B infected with T2r^ bacteriophage. J. Biol. Chem. 174: 281-293. 

DelbrOck, M., 1940 The growth of bacteriophage and lysis of the host. J. Gen. Physiol. 23: 
643-660. 
1946 Bacterial viruses or bacteriophages. Biol. Rev. Cambridge Phil. Soc. 21 : 30-40. 

Delbruck, M., and W. T. Bailey, Jr., 1946 Induced mutations in bacterial viruses. Cold 
Spring Harbor Symp. Quant. Biol. 11: 33-37. 

DELBRticK, M., and S. E. Luria, 1942 Interference between bacterial viruses. I. Interference 
between two bacterial viruses acting upon the same host, and the mechanism of virus growth. 
Arch. Biochem. 1: 111-141. 

DoERMANN, A. H., 1948 Intracellular growth of bacteriophage. Carnegie Instn. Wash. Yearb. 
(in press). 

^ While this paper was in press, one of us (Dulbecco 1949b) discovered that ultraviolet ir- 
radiated phages can be reactivated by exposure to visible light of short wave length in presence 
of bacterial cells. This "photoreactivation" differs from reactivation by multiple infection in many 
of its features. Photoreactivation does not take place to any appreciable extent under the condi- 
tions in which the experiments reported in this paper were performed. A series of experiments of 
the type exemplified in table 4, but carried out in dim yellow light — ^under conditions that com- 
pletely avoid photoreactivation — gave results undistinguishable from those of the earlier experi- 
ments, in which no precaution had been taken to control illumination. 



275 



122 S. E. LURIA AND R. DULBECCO 

DuLBECCO, R., 1949a The number of particles of bacteriophage T2 that can participate in intra- 
cellular growth. Genetics 34: (in press). 

1949b Reactivation of ultraviolet inactivated bacteriophage by visible light. Nature (in 

press). 

Foster, R. A. C, 1948 An analysis of the action of proflavine on bacteriophage growth. J. Bact. 
56: 795-809. 

Hershey, a. D., and R. Rotman, 1948 Linkage among genes controlling inhibition of lysis in 
a bacterial virus. Proc. nat. Acad. Sci. 34: 89-96. 

1949 Genetic recombination between host-range and plaque-type mutants of bacteriophage 
in single bacterial cells. Genetics 34: 44-71. 

Latarjet, R., and R. Wahl, 1945 Precisions sur I'inactivation des bacteriophages par les rayons 
ultraviolets. Ann. Inst. Pasteur 71 : 336-339. 

Lea, D. E., 1947 Actions of radiations on Hving cells. xii-|-402 pp. Cambridge: University Press. 

Lea, D. E., and M. H. Salaman, 1946 Experiments on the inactivation of bacteriophage by 
radiations, and their bearing on the nature of bacteriophage. Proc. roy. Soc, B, 133: 434-444. 

LuRiA, S. E., 1947 Reactivation of irradiated bacteriophage by transfer of self-reproducing units. 
Proc. nat. Acad. Sci. 33: 253-264. 

1948 Bacteriophage mutations and genetic interactions among bacteriophage particles in- 
side the host cell. A.A.A.S. Symposium on Genetics of Microorganisms (in press). 

Luria, S. E., and M. Delbruck, 1942 Interference between bacterial viruses. II. Interference 
between inactivated bacterial virus and active virus of the same strain and of a different 
strain. Arch. Biochem. 1 : 207-218. 

Luria, S. E., and R. Latarjet, 1947 Ultraviolet irradiation of bacteriophage during intracel- 
lular growth. J. Bact. 53: 149-163. 

Rahn, 0., 1929 The size of bacteria as the cause of the logarithmic order of death. J. gen. 
Physiol. 13: 179-205. 

APPENDIX 

ON THE RELIABILITY OF THE POISSON DISTRIBUTION AS A DISTRIBUTION OF THE 

NUMBER OF PHAGE PARTICLES INFECTING INDIVIDUAL BACTERIA 

IN A POPULATION 

R. DuLBECCO 

In calculating the average number x of phage particles adsorbed per bacteri- 
um (multiplicity of infection) from the number of uninfected bacteria, and, 
from X, the proportion of bacteria w^ith any given number k of particles, the 
assumption is made that the distribution of particles per bacterium is a 
Poisson distribution. One limitation to this assumption may arise from differ- 
ences in the surface area of individual bacteria. This limitation was analyzed 
as follows. 

In a mixture of P particles and B bacteria, each phage particle has a prob- 
ability p = c{l/B) to be adsorbed by a given bacterial cell. In actual cases 
where P and B are large and c is between 0.4 and 0.9, we have : x = cP/B. 

If c is constant for all bacteria, the distribution of ^ is a Poisson distribution; 
if not, the distribution of k will be different. We assume, in first approximation, 
that the adsorption capacity of a bacterium is proportional to its surface, and 
that the surface is proportional to the length — considering bacteria as cylinders 
with uniform diameter and negligible end surfaces. 

The distribution of bacterial lengths was obtained experimentally on stand- 
ard cultures of E. coli B containing 10^ cells per ml. Negative stains with 



276 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 



123 



nigrosin were made without fixation, and, after the slides were dried, 764 
cells were measured with an ocular micrometer (Filar). The distribution of 
lengths is given in figure 8, and will be referred to as the "B distribution." 
Measurements on five cultures gave similar distributions. 

For the purpose of obtaining the distribution of phage particles adsorbed 
per bacterial cell, or "P distribution," we shall split the bacterial population in- 



300 




MEAN= L = 25.717 MICROMETER UNITS 
VARIANCE = 109.36 (MICROMETER UNITs)^ 



0.1653 L 



T I 



150 



T — r 



200 



50 100 

MICROMETER UNITS 

Figure 8. — The distribution of bacterial lengths in standard cultures of 
Escherichia coli strain B containing 10^ cells per ml. 



to a number of subpopulations, each comprehending one of the length classes 
in figure 8, and assume that the adsorption capacity of all cells within each 
subpopulation is constant. 

By the calculation given at the end of this paper, the following properties 
of the P distribution are derived: 

a) its arithmetic mean is equal to x, that is, to the multiplicity of infection, 
as for a Poisson distribution. 

b) its variance is equal to the variance x of the Poisson distribution, plus a 
term equal to the variance of the B distribution: var(P) = a;+var(B). 

If we express the variance in terms of the mean x taken as the unit of meas- 
ure, we obtain for our experimental cultures: va,T(P) = x-\-0.1653x^. This equa- 
tion expresses the relation between the variance of the P distribution for an 
ideal population with uniform capacity of adsorption and the variance for a 
real population with a capacity of adsorption distributed with a given variance. 

The P distributions for several values of x, calculated from the actual 
B distribution, are given in table 8, together with the Poisson distributions 



277 



124 S. E. LURIA AND R. DULBECCO 

for the same values of x. It is clear that for low multiplicities of infection 
there is little difference between actual and Poisson distributions. For higher 
multiplicities, the discrepancies increase; still, for :r=1.5, they only affect 
relevantly the proportions of bacteria with 5 or more particles. 

Table 8 

The P distribution for various values of x 

"P" columns give the frequencies in the P distribution, "Poisson" columns give the fre- 
quencies in the Poisson distribution. All frequencies are multiplied by 10^ 





x=0.2 


x = l.S 


x=2.5 


x=5 


x = 7.5 x = 10 


K 




1 
















P 


POISSON 


P POISSON 


P 


POISSON 


P 


POISSON 


P 


POISSON 


P 


POISSON 



1 


82115 
15977 


81873 
16374 


24686 
32918 


22313 
33468 


10400 


8208 


1409 


674 


228 


55 


32 
218 


5 
45 


2 


1735 


1637 


23197 


25101 














765 


227 


3 


155 


109 


11658 


12551 














1908 


757 


4 


14 


6 


4751 


4707 














3739 


1892 


5 


2 




1721 


1412 














6065 


3784 


6 






611 


353 














8411 


6303 


7 






238 


76 














10218 


9009 


8 






109 


14 














11081 


11261 


9 






62 


2 














10924 


12512 


10 






38 
















9942 


12512 


11 






29 
















8489 


11375 


12 






19 
















6902 


9479 


13 






12 
















5383 


7292 


14 






8 
















4089 


5208 


15 






5 
















3036 


3472 


16 






3 
















2219 


2170 


17 






1 
















1600 


1276 


18 






1 
















1140 


709 


19 






















810 


373 


20 






















579 


187 


21-23 






















931 


147 


24-26 






















464 


11 


27-30 






















293 




31-33 






















172 




34-36 






















110 




37-39 






















67 




>39 






















305 





The calculation of x from the bacterial survival — assuming {B unin- 
fected) /5=e~^ — only gives reliable results for values of x up to about 2. For 
higher multiplicities, it leads to an underestimation of x. 

THEORY 

The bacterial population is considered as a mixture of an infinite number of 
subpopulations homogeneous in adsorption capacity. Within each subpopu- 
lation there is a Poisson distribution of phage particles per bacterium, whose 
frequency function is: 

jm = —^ (1) 

where Xj is the multiplicity of infection in thejth subpopulation. Xj is obtained 
from the B distribution of bacterial lengths: if we call Ij the length of the bac- 
teria in thejth class, we have: 



278 



ACTIVE FROM INACTIVATED BACTERIOPHAGE 125 

Xj = X— (2) 

where x = multiplicity of infection in the total population, /. = arithmetic mean 
of the B distribution (average bacterial length). 

The frequency function of the distribution of phage particles per bacterium 
in the total population (P distribution) is 

f{k) = E n^fm = Z -^—7^ (3) 

i=o j=o k ! 

where Uj is the frequency of the ^th subpopulation in the E distribution 

00 

j=o 

The Poisson distribution of phage particles per bacterium within each homo- 
geneous subpopulation of bacteria has the following first two moments: 
First moment about the origin = (;Ul')J■ = a;> 
Second moment about the mea.n = {1x2) j = Xj 
For the P distribution in the total population, remembering formula (3), we 
have a mean 

00 =0 00 00 00 00 

(mi')p = I] ^ D ^hfjik) = S %Z1 k/Ak) = Jlnj{ni)j = X) fhXj-, 

k=0 i=0 j=0 k=0 j=0 3=0 

from formula (2) we obtain: 

(mi')p = X. (4) 

To find the variance (m2)p of the P distribution, let us remember that, in 
general, 

M2 = M2' — (m/)" (5) 

where M2' is the second moment about the origin. We have then: 

CO GO CO 00 CO 

k=0 ;=0 j=0 A-=0 ;=0 

where (m2')7 is the second moment about the origin of the distribution of phage 
particles per bacterium in the jth subpopulation. But (M20y = ^y+(^j)^ (since, 
in general, M2' = M2+(mi')^)> and, for each homogeneous subpopulation, 
(m2);= itii)} = Xj. We obtain, therefore: 

(m2')p = Z ^jMi = Z «;•»;• + H nj{x,)\ (6) 

Finally, introducing the valaes obtained from (4) and (6) in formula (5) : 

00 QO 

(m2)p = E ^jXj + E ^;(^y)^ - x^- (7) 

y=o y=o 

The first term on the right side of equation (7) equals x\ the other two terms 
equal the variance of the B distribution, expressed as function of x. We obtain 
therefore, 

(m2)p = a; + var (B). 



279 



INACTIVATION OF BACTERIOPHAGES BY DECAY OF 
INCORPORATED RADIOACTIVE PHOSPHORUS* 

By GUNTHER S. STENT and CLARENCE R. FUERST| 

{From the Virus Laboratory , University of California, Berkeley) 

(Received for publication, September 29, 1954) 

It was observed by Hershey, Kamen, Kennedy, and Gest (1951) that bacteri- 
ophages are unstable if they contain radiophosphorus P^'- of high specific ac- 
tivity. From day to day, progressively decreasing fractions of such popu- 
lations of radioactive phage are still able to form plaques when plated on a 
sensitive bacterial strain, and the rate of loss of infective titer depends on the 
specific activity of the P^' assimilated. It is the purpose of this communi- 
cation to present experiments in which these observations of Hershey et al. 
have been extended to the study of the lethal effects of P^^ decay in various 
strains of bacteriophage at various temperatures and to the examination of 
some of the biological properties of the inactivated bacteriophage particles. 
Some of these experiments have already been reported in preliminary form 
(Stent, 1953 a). 

Materials and Methods 

Bacteriophages Tl, T2, T3, T5, T7, and their host, E. coli B/r, and phage X and 
its host, E. coli strain K12S, were used in this study. Strain B/r, a radiation-resistant 
mutant derived from strain B, was kindly supplied to us by Dr. Aaron Novick. 

Glycerol-casamino acid medium refers to a medium devised by Fraser and Jerrel 
(1953). H medium is a glycerol-lactate medium of the following composition per liter 
of distilled water: 1.5 gm. KCl, 5 gm. NaCl, 1 gm. NH4CI, 0.25 gm. MgS04-7H20, 
10"'* N CaCU, 0.07 M sodium lactate, 2 gm. glycerol, 0.5 gm. bacto-peptone Difco and 
0.5 gm. bacto-casamino acids Difco. H medium contains 6 mg. /liter total phosphorus, 
of which 5 mg. /liter are supplied by the casamino acids and 1 mg. /liter by the peptone. 
Control experiments show that this phosphorus is assimilated by cultures of E. coli 
neither more nor less readily than inorganic phosphate. 

The techniques described by Adams (1950) were employed for the general pro- 
cedures of bacteriophagy. 

Radiophosphorus was obtained as carrier-free H3P^^04 from the Isotope Division 
of the Atomic Energy Research Establishment, Harwell, England. Measurements 



* This investigation was supported by grants from the National Cancer Institute 
of the National Institutes of Health, Public Health Service and The Rockefeller 
Foundation. 

X Holder of a National Research Council of Canada Special Scholarship. 

Reprinted by permission of The Rockefeller Institute from 

The .Journal of General Physiology, 38 (4), 441-458, 

March 20, 1955. 

280 



442 INACTIVATION OF BACTERIOPHAGES 

of radioactivity were made on dry samples by means of an end-window GM tube, 
whose counting efficiency for P^- had been established by reference to a standard solu- 
tion of radiophosphorus supplied by the National Bureau of Standards, United States 
Department of Commerce. The specific radioactivity of the growth media was de- 
termined by radioactive counting and chemical analysis of total phosphorus in the 
case of a number of T2 lysates in order to establish the specific inactivation rate aN 
for that phage and to confirm the value obtained by Hershey et al. To conserve the 
supply of isotope, the specific activity of the growth medium in the case of the other 
phages was usually estimated only by reference to the rate of inactivation of a stock 
of T2 grown in an aliquot of the same medium. 

Bacteriophages of high specific activity were grown in the following way: A volume 
of the radioactive stock solution containing the desired amount of P^^ was evaporated 
to dryness in a boiling water bath and resuspended in 0.1 ml. of H medium. The radio- 
active growth medium was then adjusted to neutral pH and inoculated with 0.01 ml. 
of a culture of 2 X 10^ cells /ml. of B/r already in its exponential phase of growth in 
non-radioactive H medium. The growth of the radioactive culture at 37°C. was 
followed by microscopic counts in a Petroff-Hausser bacterial counting chamber. 
When the bacterial density reached 5 X 10^ cells/ml., the culture was infected with 
0.01 ml. of a stock containing 10" phages/ml. and incubated until microscopic counts 
indicated satisfactory lysis. At this point, the remainder of the 0.1 ml. culture was 
diluted into cold glycerol-casamino acid medium and assayed for its titer of infective 
phage particles. 

Experimental Results 

Rate of Inactivation. — 

Hershey et al. observed that if a stock of T2 or T4 containing P^- at high 
specific activity was assayed daily, the logarithm of the number of surviving 
phages fell linearly with the number of P^' atoms that had decayed up to the 
time of assay. The slope of this survival curve was found to be proportional to 
the specific activity of the medium in which the phages had been grown, pro- 
vided that the stock was stored in sufficiently great dilution under conditions 
in which control lysates containing an equal amount of non-incorporated 
P^2 were stable. This indicated that the inactivation of one phage particle was 
not due to the radiation emitted by the radioactivity contained in other phages 
but was the consequence of the disintegration of one of its own atoms of P'^'^ 
The rate of change in the fraction 5 of surviving phage particles with the time 
/ in days may, therefore, be expressed as 

ds/dt = -aN*\s (1) 

in which a is the fraction of the P^- disintegrations which are lethal (hereafter 
referred to as the "efficiency of killing"), A^* the number of radioactive phos- 
phorus atoms per phage particle, and X the fractional decay of P^- per day. 
Integration of (1) and substitution of more practical parameters lead to 

logio^ = -1.48 X lO-^aAoN(l - e-^') (2) 



281 



G. S. STENT AND C. R. FUERST 443 

in which Ao is the specific radioactivity (in milUcuries per miUigram of phos- 
phorus) of the growth medium and N the total number of phosphorus atoms 
per phage particle. Hence, a plot of logio5 vs. (1 — e~^^), the fraction of all P^'^ 
atoms decayed by the /'^ day, should be a straight line with slope proportional 
to ^0, the relation actually observed experimentally. 

We have studied the inactivation by P"^ decay of five virulent coliphages 
Tl, T2, T3, T5, T7, and of the temperate coliphage X. All these strains, except 
the pair T3-T7, are serologically unrelated, differ in their chemical constitution, 
morphology, genetic structure, and manner of interaction with bacterial host 
cells. Radioactive stocks of each strain were grown by the procedure indicated 
above in media ranging in specific radioactivity from 100 to 300 mc./mg. At 
these specific activities, approximately 0.03 to 0.1 per cent of all phosphorus 
atoms are present as the P^- isotope. The lysates, whose titer usually represented 
at least a thousandfold increase over the inoculum, were stored at 4°C. in 
casamino acid-glycerol medium and the number of infective centers assayed 
from day to day. The results are presented in Fig. 1 in which the logarithm 
of the fraction of the survivors in the different phage stocks is plotted against 
(1 — e~^'). It is seen that in agreement with equation (2) a straight line survival 
curve is obtained in every case. The specific death rates aN, having the di- 
mension lethal atoms per phage and obtained by dividing the observed slopes 
of the lines of Fig. 1 by -1.48 X 10-« Ao, are listed in Table I. Control ex- 
periments, not shown in Fig. 1, indicated that non-radioactive stocks of all 
six strains were stable in casamino acid-glycerol medium at 4°C. and that the 
radioactive lysates had been diluted sufficiently far to avoid inactivation by 
any external P^^. The six phages evidently fall into two classes of sensitivity 
to P^- inactivation. One class, composed of T2 and T5, is characterized by 
4.5 X 10"* lethal atoms per phage, the value already observed by Hershey 
el al. for T2 and T4. The sensitivity of the other group, comprising Tl, T3, 
T7, and X, corresponds to 1.5 X 10* lethal atoms per phage. Hence the strains 
of the second group are only one-third as sensitive to inactivation by decay 
of P^- as those of the first. 

Phosphorus Content and Efficiency of Killing. — 

The efficiency of killing per disintegration, a, may be calculated from the 
specific death rate, aN, if the number of phosphorus atoms per infective unit is 
known. The phosphorus content of each phage strain was, therefore, deter- 
mined by means of the following procedure, the results of which are listed in 
Table I. 

A stock of each phage was grown in H medium containing P^^ at a low but ac- 
curately determined specific activity. The lysate was clarified and freed of bacterial 
debris by two low speed centrifugations (10 minutes at 5,000 g) and the phage sedi- 



282 



444 



INACTIVATION OP BACTERIOPHAGES 



merited and washed three times in nutrient broth by high speed centrifugations (60 
minutes at 10,000 r.p.m. for T2, T5; 90 minutes at 15,500 r.p.m. for Tl, T3, T7). 
The number of plaque-forming units and the P^' content of the purified suspension 




8 



8 12 DAYS 



1 1 1 

A 0=270 ma/mg. 




1. 



^. 



1 1 1 

Ao=220mc./mg. 







\o 



A - 



\ 



T2 



I I 



02 0.4 0.2 0.4 j-e"^^ 

Fig. 1. P32 inactivation of Tl, T2, T3, T5, T7, and X at +4°C. Ao = specific ac- 
tivity of growth medium. 

were then assayed and the phosphorus content per infective unit calculated on the 
basis of the specific activity of the growth medium. In each case, more than 90 per 
cent of the P^^ of the purified suspension could be adsorbed specifically to sensitive 
bacterial cells, indicating that practically all the radioactivity resided in morpho- 
logically intact bacteriophage particles. The results of this analysis agree well with 
the phosphorus content of T2 determined by Hershey, Kamen, Kennedy, and Gest 



283 



G. S. STENT AND C. R. FUERST 



445 



(1951) and by Hershey and Chase (1952). The agreement is poor, however, with the 
estimations of the phosphorus contents of Tl, T2, T3, T5, and T7 by Labaw (1951) 
whose values are about twice as great as those found here. No values are listed in 
Table I for the phosphorus content of X, since it was not possible to prepare a puri- 
fied suspension of P^^-labelled X in which the bulk of the radioactivity could be ad- 
sorbed specifically to sensitive bacteria. Neither the reason for this behavior of X nor 
the nature of the non-adsorbed material has yet been discovered. 

The last column of Table I lists the efficiency of killing, a, of P^^ decay in 
each of the five strains of T phage. It is seen that in all the strains studied 
here, a is near the value 0.09 originally observed by Hershey et al.; i.e., on the 

TABLE I 

Evaluation of the Parameters of the Equation 
log,o5 = -1.48 X 10-MoaiV^(l - e'^') 
at 4°C. 



Phage 
strain 


^0 


Slope of 
death curve 


ciN 

Lethal atoms 

per phage 


P per infective 
unit 


N 
Atoms of P 
per phage 


a 




mc./mg. 






ntg. 






T2 


160* 


-10.5 


4.5 X W 


2.3 X 10-'^ 


4.5 X 10^ 


0.10 


T5 


130t 


-8.1 


4.2 X 10^ 


1.8 X 10-'" 


3.5 X lO** 


0.12 


Tl 


270t 


-7.0 


1.7 X 10^ 


0.7 X lO-" 


1.4 X 10* 


0.12 


T3 


160* 


-3.1 


1.3 X 10^ 


0.9 X 10-'" 


2 X 10^ 


0.07 


T7 


270t 


-6.4 


1.6 X 10^ 


0.9 X 10-'" 


2 X 105 


0.08 


X 


220J 


-4.8 


1.5 X W 


? 


? 


? 



* Determined radiochemically. 

J Determined by comparison with control T2 stock. 

average one of about every ten P^- disintegrations inactivates any phage 
particle in which it occurs. 

Effect of Temperature on the Efficiency a. — 

The rate of inactivation by decay of P^^ was also measured at two lower 
temperatures in the frozen state. For this purpose, aliquots of diluted radio- 
active lysates of all six phage strains were stored either at +4°C., or in the 
frozen state at — 20°C. or — 196°C. (the temperature of boiling liquid nitro- 
gen). Samples were then thawed from day to day and assayed for the fraction 
of surviving infective centers. Frozen controls with corresponding non-radio- 
active lysates showed that, depending on the strain, from 45 to 90 per cent of 
the infective centers survive freezing and thawing and that, except in the 
case of storage of T2 at — 20°C., the fraction recovered is independent of the 
length of time of storage (Sanderson, 1925; Rivers, 1927). It was found that 



284 



446 



INACTIVATION OF B.^CTERIOPHAGES 



at these lower temperatures the rate of inactivation by P^^ decay of all five 
strains was significantly reduced. Since the rate of radioactive decay is in- 
dependent of temperature, it follows that a reduction in a by the altered 
environmental conditions must be responsible for the reduced rate of bacteri- 
ophage inactivation. Table II lists the observed values of the slope of the 
inactivation curves at -f4, —20, and — 196°C. and the fractional reduction 
of a compared to its magnitude at +4°C. It is seen that radioactive decay 
proceeding at — 20°C. inactivates the phages with an efficiency of only 70 per 
cent of decay proceeding at +4°C. Lowering the temperature to — 196°C. 

TABLE II 

The Relative Efficiency of P^ Inactivation at Low Temperatures 



Phage strain 


^0 


Storage 
at +4° 


Storage at -20° 


Storage at —196° 


Slope* 


Slope* 


a(-20°) 
a(+4°) 


Slope* 


a(-196°) 
a(+4°) 


T2 

T5 

Tl 
T3 
T7 
X 


mc./mg. 

160 
130 
125 

130 
125 

270 

160 

270 

220 


-10.5 
-8.6 
-8.3 

-8.5 
-8.1 

-7.0 

-3.1 

-6.4 

-4.8 


-5.8 
-5.6 

-4.8 

-4.6 
-3.4 


0.68 
0.69 

0.69 

0.72 
0.71 


-6.8 

-5.6 

-5.7 

-4.5 
-4.6 

-3.9 

-1.6 

-3.6 

-3.3 


0.65 
0.65 
0.69 

0.53 
0.57 

0.56 

0.52 

0.56 

0.54 



* Refers to the value of -1.48 X lO'^ivxN. 

reduces the fraction of lethal disintegrations even further. At this temperature 
the efficiency of killing in Tl, T3, T5, T7, and X is only 55 per cent and in T2 
only 65 per cent of its value at +4°C. 

Since low temperatures appear to reduce the efficiency a, it seemed possible 
that radioactive decay occurring at temperatures higher than +4°C. might 
inactivate bacteriophages with greater efficiency. At elevated temperatures, 
however, bacteriophages are subject to thermal inactivation, and it is only 
possible to study the combined effects of heat inactivation and radioactive 
decay. To examine, therefore, the efficiency a at reasonably high temperatures, 
a heat-stable mutant, T53t, was first selected from our strain of T5 by the pro- 
cedure of Adams (1953). When stored in glycerol-casamino acid medium at 



285 



G. S. STENT AND C. R. FUERST 



447 



65°C. a stock of T5st loses 90 per cent of its titer in 5 hours. T5at is inacti- 
vated by P^- decay at 4°C. with the same specific death rate as the wild type 
T5. One stock of TSst was grown in H medium containing radioactive phos- 




0.02 0.04 0.06 0.08 
12 24 12 



0.02 a04 0.06 0.08 1-6" 
24 12 24 HOURS 




0.02 004 0.06 



0.02 0.04 



0.02 Q04 l-e 



Xi 



Fig. 

lysate, 



2. Inactivation of T5st at different temperatures. Filled circles, radioactive 
Ao = 300 mc./mg. Open circles, non-radioactive control lysate. 



phorus at specific activity of 300 mc./mg. (at which level 0.1 per cent of all 
phosphorus is P^-) and one in non-radioactive H medium. After dilution into 
glycerol-casamino acid medium, aliquots of both lysates were stored at 4, 
50, 55, 60, and 65°C. and assays of the number of infective centers made from 
time to time. The result of this experiment is presented in Fig. 2. It is seen 
that the rate of inactivation of the radioactive lysate is almost the same at 4, 



286 



448 



INACTIVATION OF BACTERIOPHAGES 



50, and 55°C., at which temperatures the non-radioactive control lysates 
exhibited Httle or no heat inactivation. At 60 and 65°C., however, considerable 
increases in the rate of inactivation of the radioactive TSst lysate are observed, 
at which temperatures the non-radioactive control lysate now also exhibits an 
increasing instability. Since the rate of loss of titer of the radioactive lysate 
may be presumed to be the sum of the rate of death due to heat and to radio- 
active decay, the rate of P^- inactivation can be estimated at any temperature 
by subtraction of the slope of the survival curve of the non-radioactive con- 




-196° -20" +20° +40° +60° 

TEMPERATURE DURING DECAY 

Fig. 3. The efficiency of killing, a, in T5 at different temperatures. 

trol from that of the radioactive lysate. (This subtraction of slopes is justified 
only in experiments of short duration, while (1 — e~^') is still approximated 
by \t.) The efficiency of killing a at that temperature can then be computed 
from this difference of rates by means of equation (2). The result of such 
calculations based on the slopes of Fig. 2 is presented graphically in Fig. 3, 
in which a has been plotted against the temperature of decay. It is evident 
that a increases slowly between 4 and 55°C. and begins to rise sharply after 
that point. At 65°C., a has reached the value 0.31, which means that now 
almost one in every three P^^ disintegrations is lethal to T5sf Also included 
in Fig. 3 are the results of the estimations of oc in T5 at low temperatures. 



287 



G. S. STENT AND C. R. FUERST 449 

Evidently, it is possible to effect at least a fourfold variation in a. by varying 
the temperature of storage from the lowest to the highest practicable range. 
It is to be noted that the increase in a per degree is greater between — 20 and 
+4°C. than between +4 and +50°C. This, no doubt, implies that a is afifected 
not only by the ambient thermal energy, but also by the change of phase from 
liquid to solid state. 

P^- Decay after Injection. — 

Hershey and Chase (1952) have shown that when T2 infects a sensitive 
bacterium, the phosphorus, and hence the DNA, of the bacteriophage particle 
enters the host cell, whereas the bulk of the phage protein remains outside. 
It may then be asked whether P^'^ decay can still prevent the reproduction of 
the parental phage and the ultimate emergence of infective progeny if such 
decay occurs only after the introduction of the DNA of a radioactive T2 
particle into the interior of the bacterial cell. 

In order to study the effect of P^^ decay after infection, it is necessary to 
arrest intracellular phage development reversibly for days or weeks so that 
the slow radioactive decay may proceed at an early stage of the brief 20 minute 
latent period. This can be achieved by quick-freezing the bacterial cells shortly 
after infection and storing them at — 196°C. in liquid nitrogen. As in the case 
of free phages, non-radioactive controls show that more than half of the 
infected centers survive freezing and thawing, and that the fraction recovered 
is independent of the length of storage at — 196°C. In those infected bacteria 
which survive, phage development resumes upon thawing where it had left off 
at the moment of freezing. 

A culture of strain B/r was grown in nutrient broth to a density of 10^/ml., centri- 
fuged, and resuspended in fresh broth at one-fourth of its original volume. The sus- 
pension was then infected with 3 X 10^/ml. radioactive T2 particles, containing P^^ 
at a specific activity of 88 mc./mg. Phage development was again arrested 2.5 minutes 
after infection by chilling the culture in ice. The infected bacteria were separated from 
the small fraction of unadsorbed free phage by centrifugation and resuspended in cold 
glycerol-casamino acid medium. Aliquots of 0.1 ml. of this final suspension were 
frozen and stored in liquid nitrogen. From day to day, one of the aliquots was thawed 
by addition of 1.9 ml. of warm medium and plated at once for the number of surviv- 
ing infective centers. A control culture infected with non-radioactive T2 under other- 
wise identical conditions was similarly frozen, stored, and assayed. Aliquots of the 
initial radioactive stock of free T2 and a non-radioactive control stock were also stored 
in liquid nitrogen and assayed for their survival from day to day. 

The results of this experiment are presented in Fig. 4. It is seen that in the 
population of bacteria infected for 2.5 minutes with a multiplicity of 0.075 
radioactive T2, per cell, the logarithm of the fraction of individuals capable of 
giving rise to a plaque when plated after thawing decreases linearly with 
(l — e~^0- The slope of the survival curve is about three-fourth that of the 



288 



450 



INACTIVATION OF BACTERIOPHAGES 



rate of inactivation of the free radioactive T2 stored at the same tempera- 
ture. (Neither the control culture infected with non-radioactive T2 nor the 

DAYS 
2 4 6 8 12 16 20 25 

T — T-T — I I I I I I I I I I I I I I I I Mill 



.0* 




Fig. 4. P^- inactivation of T2 (^o = 88 mc./mg.) inside infected bacteria at — 196°C. 
Filled triangles, multiplicity of infection: 0.075 (monocomplexes) . Open triangles, 
multiplicity of infection: 2.2 (multicomplexes). Filled circles, free T2. The dashed 
curve indicates the expected survival of infective centers at a multiplicity of 2.2 in 
the absence of multiplicity reactivation. 

corresponding free phage showed any significant loss of titer.) Hence P^- 
decay occurring in the DNA after it has been separated from the protein 
"coat" and exchanged its place in the phage head for the protoplasm of the 
host cell is still capable of destroying the reproductive capacity of the parent 
phage, although this inactivation now proceeds with a slightly reduced effi- 



289 



G. S. STENT AND C. R. FUERST 451 

ciency. Results similar to those presented in Fig. 4 have also been obtained 
after infection of bacteria with radioactive T3 and X phages. 

Stale of the Phage after Decay. — 

Cross-Reactivation. — The lethal damage sustained by the phage upon decay 
of one of its phosphorus atoms thus appears to prevent a step of the reproduc- 
tive cycle which occurs after the invasion of the host. In accordance with this 
view, we observed that T2 particles inactivated by P^^ decay are still adsorbed 
to bacterial cells. In fact, such phages are still able to participate in the re- 
productive processes occurring inside bacteria infected with a normal, non- 
inactivated related phage. In experiments already presented elsewhere (Stent, 
1953 h) it was found that a radioactive stock of the double mutant strain 
Tlhri could still contribute its genetic markers to the progeny of a cross with 
non-radioactive wild type T2+-1- after P^^ decay had destroyed the ability 
of the Tlhri particles to reproduce themselves in solo {cross-reactivation). 
It appeared, furthermore, that the ability of a radioactive T2 particle to donate 
either one of these two unlinked loci h and ri is destroyed separately by P^^ 
decay, each locus disappearing at about one-third the rate of the plaque- 
forming ability of the whole particle. In those infected bacteria in which only 
one of the two radioactive loci has been inactivated, the surviving locus 
appears among the progeny in nearly normal yield. Stahl (1954) also discovered 
the existence of cross-reactivation of genetic markers after inactivation of T4 
phage by P^^ decay. Stahl observed, furthermore, that the likelihood that a 
P^- disintegration prevents both of two markers from appearing among the 
progeny of a cross with an active phage is inversely related to the genetic 
linkage distance of their loci. Hence it may be inferred that the lethal damage 
of P'^ decay affects the reproduction of only part of the hereditary substance 
of the bacteriophage particle, leaving the rest intact to reproduce itself in mixed 
infection with an active phage. 

Multiplicity Reactivation.- — The presence of an active phage particle in the 
same bacterial cell, however, appears to be necessary for the survival of the 
undamaged parts of a P^^-inactivated T2 phage. Contrary to ultraviolet-in- 
activated T2 (Luria, 1947), infection of one bacterium by several P^'-in- 
activated particles does not lead to the production of active phage {multi- 
plicity reactivation) . 

In order to test for multiplicity reactivation following P^- decay, the stock of radio- 
active T2 employed in the experiment presented in Fig. 4 was used to infect B/r 
bacteria at a multiplicity of 2.2 phage particles per cell. As in the low multiplicity 
experiment of Fig. 4, the mixture of bacteria and radioactive phage was incubated at 
37°C. for 2.5 minutes before being frozen, stored at — 196°C., and assayed for surviv- 
ing infective centers from day to day. At a multiplicity of infection of 2.2, the frac- 
tion of all infected bacteria to which two or more phages are adsorbed (multicom- 
plexes) is 0.73. Hence if two or more T2 particles were able to cooperate in the 



290 



452 



INACTIVATION OF B.\CTERIOPHAGES 



production of active progeny after each individual had already sustained a "lethal" P^- 
disintegration, then the rate of inactivation of 0.73 of the plaque formers in this ex- 
periment should have been significantly reduced over the rate of inactivation of singly 
infected cells. If, on the other hand, the plaque-forming ability of a multiply infected 
cell is destroyed as soon as each of the infecting particles has been inactivated by P^'^ 
decay, then the infective centers in this experiment should have disappeared with the 
"multiple hit" kinetics indicated in Fig. 4 by a dashed curve. The result of this ex- 
periment is also shown in Fig. 4. It is seen that inactivation of multicomplexes pro- 
ceeds at roughly the same rate as inactivation of singly infected bacteria, indicating 
the absence of any appreciable multiplicity reactivation. Experiments in which P'^ 
decay was first allowed to take place in free T2 and in which bacteria were then mul- 
tiply infected with the inactivated phages likewise failed to reveal any multiplicity 
reactivation. 

Latent Period of Survivors. — Since the efficiency of killing, a, is less than 
0.1 at low temperatures, it is apparent that after an amount of decay which 

TABLE III 
Pholoreaclivation of T2 



Treatment of T2 


Assayed in dark 


Assayed in light 




Titer 


Titer 


Before P'^ decay 
After P32 decay 

Before ultraviolet irradiation 
After ultraviolet irradiation 


1.7 X 108 
1.7 X 10^ 

2.5 X 10' 
1.4 X 10* 


1.7 X 108 

1.3 X 10^ 

2.5 X 10' 
1.3 X 10' 



leaves only a small fraction of the initial phage population still active has 
taken place under these conditions there have occurred many non-lethal 
P^2 disintegrations in the survivors. In the case of T2, these survivors, how- 
ever, exhibit no evident effects of this non-lethal decay and reproduce with 
normal latent period and burst size. This is in contrast to the survivors of 
ultraviolet light irradiation whose multiplication is significantly retarded 
(Luria, 1944). 

Photoreactivation. — T2 bacteriophages inactivated by ultraviolet light can 
be "photoreactivated" by exposure of bacteria infected with such phages to 
visible light (Dulbecco, 1949). To examine whether phage inactivated by 
decay of incorporated P''" could be similarly reactivated by light, assays were 
made of a radioactive T2 stock before and after decay to 0.0001 of the initial 
titer, incubating the assay plates either in the dark or under a strong fluo- 
rescent light. A non-radioactive control stock of T2 was inactivated with ultra- 
violet light to a survival of 0.000056 and similarly assayed in dark and light. 
The result of this experiment is presented in Table III, in which it may be seen 



291 



G. S. STENT AND C. R. FUERST 453 

that no photoreactivation of the P-^^-inactivated T2 took place, although the 
titer of the ultraviolet-inactivated control was raised by nearly a factor of 100 
by exposure to visible light. 

DISCUSSION 

Cause of Death. — 

An atom of P^^ decays into the stable isotope of sulfur, S^^, upon ejection 
of a beta electron-neutrino pair of total kinetic energy 1.7 mev. The beta 
particle produces ionizations along its path, which are capable of damaging 
biological materials in a way similar to x-rays. Hershey, Kamen, Kennedy, 
and Gest, however, showed by means of calculations based on the volume of 
the T2 particle, the density of ionizations along the beta track and the known 
efficiency of killing per x-ray ionization, or by reconstruction experiments 
in which non-radioactive phage particles were irradiated with beta particles 
emitted by external, non-incorporated P^^ atoms, that beta particle ioniza- 
tions could not be the principal cause of the inactivation of radioactive bacteri- 
ophage particles. Hershey et al. concluded, rather, that a short range conse- 
quence of the nuclear reaction, e.g. the recoil sustained by the disintegrating 
nucleus upon ejection of beta electron and neutrino, or the transmutation 
of phosphorus into sulfur, was responsible for death. The present finding that 
the sensitivity of radioactive phages to P'^ decay is reduced only slightly 
after infection supports this view. For, it appears likely that the state of ag- 
gregation of the phage DNA is more compact in the phage head than in the 
protoplasm of the host cell (Watanabe, Stent, and Schachman, 1954). Hence 
the chance of irradiation of one part of the phage DNA by distant P^^ atoms of 
another would have been seriously reduced once infection was under way. 

Efficiency of Killing. — 

Hershey et al. suggested that the fact that only one P^^ disintegration in 
about ten was lethal to T2 or T4 might reflect a division of the phage DNA 
into 10 per cent "essential" and 90 per cent "non-essential" structures. Under 
this view, any P^^ disintegration in the former would be surely lethal and any 
in the latter generally harmless. The present finding that a is nearly the same 
in various phage strains of greatly different size, morphology, and biological 
properties makes this hypothesis less likely. The dependence of a on tem- 
perature, furthermore, excludes the possibility that the anatomy of the phage 
is the sole factor responsible for the efficiency of killing. It seems, rather, 
that a must at least in part reflect some structural aspect of the DNA mole- 
cule, the substance whose function is presumably destroyed by the decay of 
its radioactive P^^ atoms. 

The lethal effects of P^^ decay can perhaps be best understood in terms of 
the macromolecular structure of DNA, recently uncovered by Watson and 



292 



454 



INACTIVATION OF BACTERIOPHAGES 



Crick (1953), of which a schematic diagram is presented in Fig. 5. This struc- 
ture reveals DNA as a double helix composed of two intertwined polynucleotide 
chains of opposite polarity held together laterally by specific hydrogen bonds 
between purine and pyrimidine bases of opposite strands. The radioactive 
P^2 atoms are located in the diester bonds responsible for the continuity of 



interru 
ch 




on-lethol 
decay 



nterrupted 
ain 



Fig. 5. Schema of the Watson-Crick structure of DNA. The two ribbons sym- 
bolize the two phosphate- sugar chains, and the horizontal rods represent the pairs 
of bases holding the chains together through a pair of hydrogen bonds. The breaks 
in the ribbons indicate the spontaneous interruptions of the polynucleotide chains 
proposed by Dekker and Schachman. 

the polynucleotide chains. It appears almost inevitable that every ester linkage 
is destroyed upon decay of its radioactive phosphorus atom. First of all, the 
maximum recoil sustained by the phosphorus nucleus is of the order of 80 ev. 
(the average value being somewhat lower owing to the random orientation 
of neutrino and beta electron), whereas the energy of theP-0 bond holding the 
atom in place is less than 5 ev. The ester bond is, therefore, probably broken 
by the Szilard-Chalmers reaction {cf. Libby, 1947). Secondly, even if the 



293 



G. S. STENT AND C. R. FUERST 455 

recoil does not rupture the phosphate ester linkage, i.e. if the phosphorus 
nucleus remained in place after all, then the two deoxyribose residues are forth- 
with linked by a sulfate diester, which should undergo spontaneous hydrolysis 
in aqueous medium (Kremann, 1907). Inspection of the structure shown in 
Fig. 5 indicates, however, that breakage of one ester link would not neces- 
sarily lead to the disruption of the DNA molecule, since the multitude of 
hydrogen bonds still hold the two sister strands together. This has recently 
been pointed out by Dekker and Schachman (1954), who propose on the 
basis of physicochemical evidence that the polynucleotide strands of "native" 
DNA are not actually continuous throughout the length of the macromolecule 
but are already interrupted in such a fashion that on the average one out 
of twenty to fifty phosphate links is singly instead of doubly esterified, as 
indicated in Fig. 5. Thus, if there already exist spontaneous breaks within 
intact DNA, it is not unreasonable to suppose that the low efficiency of killing 
per P^- disintegration means that the DNA molecule can continue to function 
even after a few additional interruptions of the polynucleotide chains have 
been generated by radioactive decay. 

An event secondary to the disruption of the phosphate diester must then 
attend the lethal fraction a of P^- disintegrations. The most reasonable hy- 
pothesis would appear to be that inactivation is caused by a complete cut of 
the DNA double helix. One way in which this could occur is that enough energy 
liberated by the decaying P^- atom has been transmitted by a sequence of elastic 
collisions to the other strand to also cause a break there. Another possibility, in 
view of the proposal by Dekker and Schachman, would be that the lethal de- 
cay takes place in an atom situated nearly in apposition to one of the few in- 
complete ester links on the other strand. In either case, a complete cut results 
because few or no hydrogen bonds remain between the spots where both sister 
strands are broken to oppose the dissociation of the macromolecule into two 
smaller pieces. The effect of heat on the efficiency a is readily explained in 
terms of this hypothesis. The rapid rise of a above 55°C. must be due to the 
dissociation of the hydrogen bonds at these temperatures {cf. Dekker and 
Schachman), thus causing less and less resistance to separation of the two 
strands by the energy of the radioactive disintegration. A greater and greater 
fraction of the P^- decays can, therefore, result in a complete cut of the double 
helix. The effect of freezing and of low temperatures on reducing a might be 
explained by the increase of viscosity of the medium in which the two pieces 
involved in the break have to move; i.e., that when the DNA is embedded in 
ice there exists a greater chance that the energy of the P^' transmutation has 
already been dissipated before the cut has actually taken place. 

Action of Ionizing Radiations. — 

It would be possible, though technically rather difficult, to ascertain whether, 
in agreement with the hypothesis just proposed, decay of incorporated P^^ 



294 



456 INACTIVATION OF BACTERIOPHAGES 

actually depolymerizes highly radioactive DNA molecules with an efficiency 
similar to a. It is known, however, that x-rays and other ionizing radiations do 
break down DNA to random fragments of progressively smaller molecular 
weight at doses comparable to those necessary for the "direct" inactivation of 
bacteriophages (Taylor, Greenstein, and Hollaender, 1948; Conway, Gilbert, 
and Butler, 1950). Hence it is not unlikely that the lethal effect of x-ray ioniza- 
tions inside the phage particle is also one of cutting DNA molecules, similar 
to that postulated above for P^'- decay. Two sets of facts would appear to make 
this comparison useful : 

(c) The efficiency of killing per x-ray ionization inside the volume of the 
phage particle is only of the order of 0.05 in the bacteriophage strains studied 
here (Watson, 1950); i.e., similar in magnitude to a. The energy released by 
each x-ray ionization is thought to be 32 ev., i.e. similar in magnitude to that 
of the P^^' recoil, and to be confined to a radius of a few Angstrom units (Lea, 
1947). (The average energy available locally may actually be either more or 
less than 32 ev. because, on one hand, the ionizations tend to occur in clusters 
but, on the other hand, their energy has been determined only in air and not 
in a condensed phase.) Since the two polynucleotide chains of the DNA macro- 
molecules are separated by at least 10 A (Watson and Crick, 1953), it would 
appear possible that many of the ionizations, like many of the P^' disintegra- 
tions, damage only one of the strands without causing a complete rupture of 
the double helix. 

(b) The x-ray sensitivity of Tl depends on temperature very much like 
a. At temperatures below freezing, the rate of inactivation by x-rays is only 
65 per cent of that just above freezing (Bachofer et al., 1953). At higher tem- 
peratures, the sensitivity first remains relatively constant and then increases 
sharply above 50°C., reaching a sixfold greater value at 60°C. (Adams and 
Pollard, 1952). These observations had already suggested to Adams and 
Pollard that the weakening of secondary, interchain bonds by heat at the 
moment of the x-ray ionization might be responsible for increasing the chance 
of causing lethal damage at higher temperatures. As in the case of P'*- decay, 
it is apparent that the greater the extent to which the hydrogen bonds of the 
DNA macromolecule are dissociated, the more likely will a cut of the double 
helix result from an energetic rupture of a single polynucleotide strand. 

SUMMARY 

The inactivation of the phages Tl, T2, T3, T5, T7, and X by decay of in- 
corporated P^- has been studied. It was found that these phages fall into two 
classes of sensitivity to P^- decay: at the same specific activity of P^- in their 
deoxyribonucleic acid (DNA), T2 and T5 are inactivated three times as rapidly 
as Tl, T3, T7, and X. Since the strains of the first class were found to contain 



295 



G. S. STENT AND C. R. FUERST 457 

about three times as much total phosphorus per phage particle as those of the 
second, it appears that the fraction of all P^- disintegrations which are lethal 
is very nearly the same in all the strains. This fraction a depends on the tem- 
perature at which decay is allowed to proceed, being 0.05 at — 196°C., 0.1 at 
+4°C., and 0.3 at 65°C. 

Decay of P^^ taking place only after the penetration of the DNA of a radio- 
active phage particle into the interior of the bacterial cell can still prevent the 
reproduction of the parental phage, albeit inactivation now proceeds at a 
slightly reduced rate. T2 phages inactivated by decay of P'^ can be cross- 
reactivated; i.e., donate some of their genetic characters to the progeny of a 
mixed infection with a non-radioactive phage. They do not, however, exhibit 
any multiplicity reactivation or photoreactivation. 

The fact that at low temperatures less than one-tenth of the P^^ disintegra- 
tions are lethal to the phage particle and the dependence of the fraction of 
lethal disintegrations on temperature can be accounted for by the double 
stranded structure of the DNA macromolecule. 

REFERENCES 

Adams, M. H., Methods of Medical Research, (I.H. Comroe, editor), Chicago, The 

Year Book Publishers, 1950, 2, 1. 
Adams, M. H., Ann. Inst. Pasteur, 1953, 84, 1. 

Adams, W. R., and Pollard, E. C, Arch. Biocheni. and Biophysic, 1952, 36, 311. 
Bachofer, C. S., Ehret, C. F., Mayer, S., and Powers, E. L., Proc. Nat. Acad. Sc, 

1953, 39, 744. 
Conway, B. E., Gilbert, L., and Butler, J. A. V., J. Chem. Soc, 1950, 3421. 
Dekker, C. A., and Schachman, H. K., Proc. Nat. Acad. Sc, 1954, 40, 894. 
Dulbecco, R., /. Bact., 1949, 59, 329. 

Eraser, D., and Jerrel, E. A., /. Biol. Chem., 1953, 205, 291. 
Hershey, A. D., and Chase, M., /. Gen. Physiol., 1952, 36, 39. 
Hershey, A. D., Kamen, M. D., Kennedy, J. W., and Gest, H., J. Gen. Physiol., 1951, 

34, 305. 
Kremann, R., Monatsh. Chem., 1907, 28, 13. 
Labaw, L. W., J. Bact., 1951, 62, 169. 
Lea, D. E., Action of Radiations on Living Cells, Cambridge, University Press, 

1947. 
Libby, W. F., /. Am. Chem. Soc, 1947, 69, 2523. 
Luria, S. E., Proc Nat. Acad. Sc, 1944, 30, 393. 
Luria, S. E., Proc Nat. Acad. Sc, 1947, 33, 253. 
Rivers, T. M., J. Exp. Med., 1927, 45, 11. 
Sanderson, E. S., Science, 1925, 62, 377. 
Stahl, F. W., in a discussion of a paper by Doermann, Chase, and Stahl, /. Cell, and 

Comp. Physiol., 1954, in press. 
Stent, G. S., Cold Spring Harbor Symp. Quant. Biol., 1953 a, 18, 255. 



296 



458 INACTIVATION OF BACTERIOPHAGES 

Stent, G. S., Proc. Nat. Acad. Sc, 1953 b, 39, 1234. 

Taylor, B., Greenstein, J. P., and Hollaender, A., Arch. Biochem., 1948, 16, 19. 

Watanabe, I., Stent, G. S., and Schachman, H. K., Biochim. et Biophysic. Ada, 1954, 

15, 38. 
Watson, J. D., /. BacL, 1950, 60, 697. 
Watson, J. D., and Crick, F. H. C, Nature, 1953, 171, 737; Cold Spring Harbor Symp. 

Quant. Biol, 1953, 18, 123. 



297 



RESISTANCE TO ULTRAVIOLET LIGHT AS AN INDEX TO 
THE REPRODUCTION OF BACTERIOPHAGE 

S. BENZERi 

Calijornia Institute of Technology, Pasadena, California 
Received for publication July 18, 1951 

Infection of a susceptible bacterium by a single phage particle initiates a 
series of events climaxed, after a time called the latent period, by bursting of 
the cell and the release of a number (burst size) of replicas of the initial phage. 
We are here concerned primarily with the intervening process of phage replica- 
tion which takes place behind the cloak of the cell wall. By prematurely dis- 
rupting infected cells, Doermann (1948) found that infective phage replicas 
are already present well before the time at which the bacterium bursts, i.e., 
about two-thirds of the way through the latent period. At earlier times, however, 
no plaque-forming particles are recovered, not even the initial phage. Our atten- 
tion is, therefore, focused upon this "dark" period, during which the infecting 
phage must undergo some modification, and the key processes of phage repro- 
duction come to pass. 

Luria and Latarjet (1947) conceived the following experiment in an attempt 
to use target theory for an analysis of the intracellular developments. It had 
been shown by Anderson (1948) that a bacterium {Escherichia coli, strain B) 
could be subjected to rather heavy doses of ultraviolet light and still survive in 
its ability to support the growth of phage T2. Thus, if one were to infect cells 
of strain B with single particles of T2 and irradiate the phage-bacterium com- 
plexes, the survival of infectivity (the ability to release at least one phage particle, 
thereby forming a plaque) of the complex should be determined by the survival 
of the phage part. If the irradiation is done immediately after infection, one 
should obtain the same survival curve for complexes as for the free phage ir- 
radiated before addition to bacteria, i.e., an approximately exponential or "one- 
hit" curve. If complexes are allowed to develop to the point where several intra- 
cellular phage particles are present, the inactivation of the complex requires at 
least one "hit" in each phage, and a multiple-target survival curve should be 
obtained. The set of curves for samples irradiated at different stages in the latent 
period would be expected to resemble the theoretical curves of figure 1. These 
multiple-target curves, plotted on a semilogarithmic graph, are characterized 
by asymptotes of constant slope equal to that for the single phage particle. The 
intercept of the asymptote, extrapolated to zero dose, corresponds to the loga- 
rithm of the number of targets. 

^ On leave of absence from the Department of Physics, Purdue University. This investi- 
gation was conducted in part at Oak Ridge National Laboratory and continued while the 
author was at the California Institute of Technology as a Postdoctoral Fellow of the Atomic 
Energy Commission, and a Fellow in Cancer Research of the American Cancer Society, 
recommended by the Committee on Growth. 

Present Address: Institut Pasteur, Paris 15, France. 

Reprinted by permission of the author and the Williams and 

Wilkins Co., from the Journal of Bacteriology, 63 (1), 59-72, 

January, 1952. 

298 



60 



S. BENZER 



[vol. 63 



In experiments with E. coli, strain B, and phage T2, Luria and Latarjet found 
that immediately after infection the survival curve agreed with that of free 
phage. However, at later times, instead of showing a progressive increase in 
multiplicity with constant slope, the curves during the first half of the latent 
period showed a progressive decrease in slope (i.e., a decrease in sensitivity to 
ultraviolet) while remaining essentially exponential in character. At mid-latent 
period the curves became multiple-target in character, and thereafter the 
sensitivity increased again. Since the results did not resemble the family of curves 




Figure 1. Theoretical survival curves for complexes (after Luria and Latarjet, 1947). 

For an individual target the survival y is given by e~^ , where D is the dose in arbitrary 
units. 

For a complex containing n identical, independent targets, the survival of any one of 
which is sufficient for survival of the complex, y = 1 — {1 — e~^)". 

in figure 1, it was not possible to perform a target-theory determination of the 
number of intracellular phage particles. Latarjet (1948) repeated these studies, 
using X-rays, and observed changes which were similar in character although 
differing in degree. 

Further investigation of this problem was suggested by the discovery of 
"multiplicity reactivation" (Luria, 1947; Luria and Dulbecco, 1949). In this 
phenomenon, two ultraviolet inactivated phage particles, when infecting the 
same cell, can combine their resources, leading to the production of active phage. 



299 



1952] ULTRAVIOLET LIGHT INDEX IN BACTERIOPHAGE 61 

This result may also be stated in the following way : complexes formed by infec- 
tion of a cell with two phage particles do not have a survival curve corresponding 
to the two-target curve of figure 1 ; at any given dose of ultraviolet the probability 
of the complex being infective is greater than given by that curve. This effect 
occurs to a marked degree with phage T2 and would be expected to cause anom- 
alous results in an experiment of the Luria-Latarjet type, where intracellular 
multiplication is going on. However, there are other phages, T7 for instance, 
with which multiplicity reactivation does not occur, and it seemed of interest 
to extend the experiments to such a phage. 

Another possible cause of anomalous results is "photoreactivation" of phage 
(Dulbecco, 1950). Ultraviolet inactivated phage particles may be reactivated, 
after adsorption to a sensitive bacterium, by exposure to white light. Thus, we 
may expect the infectivity of a phage-bacterium complex to have a higher 
resistance to ultraviolet if exposed to light (after ultraviolet irradiation) than 
if kept in the dark. Precautions are therefore necessary in order to avoid this 
effect. 

MATERIALS 

Phages: T7, T2, and T2r, prepared from lysates in broth, purified by centrif- 
ugation, and resuspended in buffer. 

Bacterium: Escherichia coli, strain B, grown in broth. 

Growth medium (broth): bacto-tryptone, 1 per cent plus 0.1 m NaCl. 

Buffer: 1/15 m phosphate buffer, pH 7, plus 0.1 m NaCl, plus lO"' m MgS04. 

METHODS 

A Luria-Latarjet experiment involves the following steps: 

1. Phage particles are added to a suspension of bacteria and time is allowed 
for infection of the cells to occur, then unadsorbed phage is eliminated. 

2. The complexes are allowed to develop and samples are removed at various 
stages of the latent period. 

3. Each sample is exposed to several doses of ultraviolet. 

4. Aliquots are plated to determine the fraction of infective centers surviving 
each dose. These operations must be completed before the end of the latent 
period. Furthermore, since the radiation resistance of the infective centers 
changes rapidly with time, it is essential for accurate results that growth start 
almost simultaneously in all cells and that it be halted during irradiation. 

In an attempt to best satisfy these requirements, the following procedure is 
used : 

1. Adsorption of phage without growth. The bacteria are prepared from an 
aerated broth culture in the logarithmic phase (1 X 10^ cells per ml). The cells 
are centrifuged, resuspended in buffer, centrifuged again, resuspended in buffer 
at a concentration of 1 X 10^ per ml, and aerated by bubbling at 37 C for one 
hour in order to exhaust intracellular nutrients and bring the bacteria to a starved 
condition. A purified suspension of phage particles in buffer is then added. 
Under these conditions adsorption takes place, but no lysis or phage liberation 



300 



62 



S. BENZER 



[vol. 63 



(Dulbecco, 1950). However, the absence of phage hberation is not sufficient to 
exclude the possibihty that intracellular growth progresses to a fairly advanced 
intermediate stage without reaching completion. 

We can test for this by making use of the large change in resistance to ultra- 
violet of T2r complexes. If intracellular development proceeds, the resistance 
should increase with time. In figure 2, it can be seen that so long as the infected 
cells are kept in buffer and no nutrient is added, the resistance remains constant. 
The value of the resistance is only slightly higher than that of the free phage 




f— in broth ot 37' 



T2I 



f 



buffer ot 37 



50 



60 



10 20 30 40 

TIME IN MINUTES 
Figure 2. Effect of addition of nutrient upon the radiation resistance of T2r complexes 
formed in buffer. The ordinate is obtained by exposing a sample of complexes to ultraviolet 
for 55 seconds and determining the fraction which survives. (A) No nutrient added; (B) 
broth added at time zero. 



(see later). Thus, the intracellular development does not progress beyond a 
very early stage. Even if the adsorption period extends over many minutes, 
each complex is arrested in its development, and adsorption is effectively simul- 
taneous. 

In experiments with bacteria grown in synthetic medium (instead of broth) there was a 
slow increase of resistance, even in buffer, as if considerable intracellular nutrient reserves 
remained even after starving for an hour. This resistance reached a maximum value corre- 
sponding to that attained after several minutes of growth in the presence of ample extra- 
cellular nutrient. 



301 



1952] 



ULTRAVIOLET LIGHT INDEX IN BACTERIOPHAGE 



63 



T2r is adsorbed to the extent of over 90 per cent in 10 minutes in buffer. At 
the same buffer concentration, T7 adsorption is extremely slow, but the rate is 
greatly increased by using buffer diluted 10 times with distilled water (Watson, 
personal communication). Therefore, throughout the experiments with T7 diluted 
buffer was used. The number of infective centers obtained (with either phage) 
was usually found to be less, by one-third to one-half, than the number of phages 
adsorbed. The cause of this "abortive adsorption" is not understood. To elimi- 
nate unadsorbed phage, the cells are washed by centrifugation and resuspended 
in fresh buffer. 



100- 




> 

u 

"^0 10 20 30 40 

TIME IN MINUTES AFTER ADDING BROTH 
Figure 3. One-step growth curve at 37 C for T2r complexes formed in buffer. Broth 
added at time zero. "Infective center" signifies a plaque-forming unit, either an infected 
cell or a free phage particle. 



2. Growth. To the suspension of infected cells in buffer at 37 C, an equal volume 
of broth at the same temperature is added, and time is reckoned from this 
moment. In the case of phage T2r the radiation resistance promptly begins to 
rise (figure 2). From the one step growth curve in figure 3, it may be seen that 
the (minimum) latent period for T2r is 19 minutes, and the average burst size 
is 60. For T7, the latent period is 12 minutes, and the burst size is 150. To stop 
development at any chosen time during the latent period, a sample is removed 
and diluted rapidly (by blowing out of a pipette) into buffer chilled in an ice 
bath. The chilling brings the change in radiation resistance to an immediate 



302 



64 s. BENZER [vol. 63 

halt; the resistance remains constant for hours provided the sample is kept 
chilled. The pipettes used for addition of broth and removal of samples are 
previously equilibrated at 37 C in an incubator. 

In this manner, growth may be started and stopped in all cells simultaneously, 
and the timing controlled to within a few seconds. Furthermore, many samples 
may be taken at close intervals within the same growth experiment and the 
irradiation conducted afterwards at leisure. 

For samples chilled very close to the end of the latent period (after 16 minutes 
for T2r or 9 minutes for T7) lysis is not prevented by chilling. After a delay, 
there appears a gradual increase in plaque count, presumably due to slow lysis 
of cells containing completed phage particles. 

3. Irradiation. One or two ml of the suspension to be irradiated are placed in 
a shallow watch glass. The suspension is transparent to ultraviolet by virtue of 
the large dilution (100 X) of the broth. Ice in a small dish is used to chill the 
suspension from below (figure 4). This prevents growth during the irradiation. 
Chilling also serves to reduce greatly the rate of photoreactivation (Dulbecco, 
1950) which might otherwise be caused by the visible light emitted by the 
ultraviolet lamp. 




Figure 4- Arrangement for chilling sample during irradiation. 

Ultraviolet is supplied by a 15 watt "germicidal lamp" (General Electric 
Company). The energy emitted in the 2537 A line accounts for almost all the 
antiphage activity. The intensity of ultraviolet is such that T2r survives to the 
extent of 10~- after an exposure of 40 seconds. This same intensity is used 
throughout the experiments, and the doses are, therefore, given in units of sec- 
onds. All manipulations after irradiation are conducted in dim yellow light to 
minimize the possibility of photoreactivation. 

4. Plating. Aliquots of samples subjected to various doses of ultraviolet are 
plated in a top layer of soft agar, seeded with unirradiated B, on broth agar 
plates. The resultant plaques are counted to determine the fraction of complexes 
whose infectivity has survived the irradiation. 

Note on multiflicity of infection. It is essential that the proportion of phage 
particles to bacteria be kept quite small in order that very few multicomplexes 
(i.e., bacteria infected with more than one phage particle) be formed. This is 
particularly important in the case of T2, where the phage exhibits multiplicity 
reactivation and the multicomplexes are much more resistant to ultraviolet 
than monocomplexes. Thus, if one assumes a Poisson distribution of phages 
per bacterium, and a survival value of 10"^ is to be accurate within 10 per cent, 
the average multiplicity of infection must be 2 X 10~^ or less. In order to have 



303 



1952] ULTRAVIOLET LIGHT INDEX IN BACTERIOPHAGE 65 

such a low multiplicity of infection and still have a measurable concentration 
of survivors after growth, dilution, and irradiation, it is necessary to start with 
a high concentration of bacteria (1 X 10^ per ml). Concentrations either half 
or twice this value were found to give the same results. 

PRELIMINARY CONSIDERATIONS 

1. Ability of irradiated bacteria to support phage growth. It is necessary to know 
the validity, under the conditions used, of the basic assumption that the irradia- 
tion of the bacteria does not affect their ability to support phage growth. For 
this reason Anderson's experiments were repeated, using phage T2r. Starved 
bacteria, prepared as previously described, were irradiated in buffer before addi- 
tion of phage. The number of infective centers produced per adsorbed phage was 
determined by plating and compared to the result with unirradiated bacteria. 
At the intensity of ultraviolet here used, 4,000 seconds of irradiation are required 
to reduce to one-half the number of bacteria capable of liberating phage after 
infection with T2r. Since the largest doses used in irradiating infected cells were 
1 ,000 seconds, inactivation of the bacterial part of the complex should have had 
a negligible effect upon the survival curve. It is conceivable, however, that the 
sensitivity of the bacterial part of the complex does not remain unchanged 
throughout the latent period. Indeed, it becomes difficult to separate the complex 
into phage and bacterial components once growth has started. 

For T7, it is likewise found that the highest dose used (350 seconds) has 
negligible effect upon the ability of the bacteria to yield phage after infection. 

2. Effect of irradiation of complexes upon the latent period and burst size of the 
survivors. It is found that a progressive lengthening of the latent period and 
decrease in burst size are produced by increasing doses of ultraviolet. For T2 
monocomplexes irradiated at mid-latent period with a dose such that 5 per cent 
survive to form visible plaques, the latent period of the survivors is doubled 
and the burst size is reduced to 10 per cent of normal. This effect results in a 
smaller and more variable size of the plaques formed by surviving monocom- 
plexes. The degree of the effect appears to be determined primarily by the dose 
of ultraviolet rather than the percentage inactivation of the complexes. There- 
fore, it is most apparent at times during the latent period when the largest doses 
are required for obtaining survival curves. In plotting survival curves and at- 
tempting to analyze them by target theory, we are assuming that inactivation 
of a complex is an all-or-none phenomenon. It must be realized that the delay 
in lysis and reduction in burst size of the surviving infective centers could also 
lead to failure of some of them to produce visible plaques. However, for the 
doses used in this paper, the plaques observed with T2r and T7 do not taper 
gradually down to zero in diameter; probably few infective centers fail to be 
counted for this reason. 

RESULTS 

Experiments with T2r. T2 (used by Luria and Latarjet) and its mutant T2r, 
which has the same sensitivity to ultraviolet as T2, gave similar results. It is 



304 



66 



S. BENZER 



[vol. 63 



preferable to work with T2r because it produces larger plaques, thereby obviating 
in some measure the difficulty created by the decrease in plaque size at large 
doses. 

In figure 5 a complete set of curves for T2r (latent period 19 minutes) is 
given. These qualitatively confirm the observations of Luria and Latarjet with 
T2. The free phage does not have a strictly exponential survival curve, but the 
changes in resistance during the latent period are so large that this may be 
ignored for our purposes. At t = (i.e., for a sample irradiated before the addi- 
tion of nutrient) the resistance of monocomplexes is slightly higher (by 20 per 
cent) than that of free phage. This may be partially due to a small amount of 
development of the phage which can take place in buffer. 





400 500 
SECONDS 



Figure 5. Survival curves for T2r complexes irradiated at different times during the 
latent period. Each curve is marked with the time in minutes after adding broth to the 
complexes formed in buffer. All data are from the same growth experiment at 37 C (latent 
period = 19 minutes). Average multiplicity of infection = 2 X 10"'. 

As development proceeds, the resistance of all complexes increases. An es- 
pecially rapid rise occurs between 6 and 8 minutes, the curves still remaining 
essentially exponential. The smallness of the bend (at large doses) in the survival 
curve taken at 7 minutes is very significant, since it implies that, at least up to 
this point (one-third of the way into the latent period), development has pro- 
gressed rather uniformly in all infected cells. If, at 7 minutes, 10 per cent of the 
infective centers had progressed to a stage corresponding to 8 minutes (by 
virtue of 15 per cent more rapid development), the bend in the 7 minute curve 
would have been greater than observed. 

At 9 or 10 minutes the resistance reaches a maximum phase, and the curve 
has now become "multiple-target" in character. In figure 6 data are given which 
were obtained at 10 minutes. The points fit very well a theoretical curve for a 



305 



1952] 



ULTRAVIOLET LIGHT INDEX IN BACTERIOPHAGE 



67 



double target. However, the fit may be fortuitous, since slight inhomogeneities 
in the stage of development are to be expected. At the time of maximum resis- 
tance, those complexes which are either slightly ahead or behind in stage of de- 
velopment have lower resistances, causing a distortion of the true curve at the 
moment of maximum resistance. Thereafter, the curve retains a multiple-target 
shape, but the resistance of the individual targets, as judged by the final slope, 
decreases progressively as the end of the latent period is approached. 



I 9-^ 
0.9- ~ 
0.8- 
07- 



O.Gr 
0.6 

0.4 
03 

0.2 



o 0.1 
,0.09 
<0.08 
>0.07 
>0.06 

i>^0.05 
0.04 



0.03 



theoreticol ^ 
curve 



600 800 1000 
IN SECONDS 



200 400 

UV DOSE 

Figure 6. Circles: Survival data obtained for T2r complexes irradiated at the phase of 
highest resistance to ultraviolet (i = 10 minutes). 

Dashed curve: Theoretical survival curve for a double-target complex. 



The late curves have slight "tails" which may be due to a small fraction of 
cells in which the development is retarded. 

In some experiments, peculiar composite survival curves were obtained as though the 
development in about half the cells was arrested at a stage corresponding to around 8 
minutes, while the remaining cells continued normally. It has not been found possible to 
clarify the conditions leading to this result. 

According to Doermann's findings, there may be a few infective phage particles 
per cell at the latest time here studied (15 minutes). These would presumably 
have the resistance of free phages. If survival as an infective center depends 
upon survival of either the complex or at least one completed phage, the effect 



306 



68 



S. BENZER 



[vol. G3 



of the phages upon the observed survival cun^e will be negligible, since at 
15 minutes the complex has a much higher survival than have several free phages. 
By reference to figure 1 it can be seen that even a cluster of 100 phages would be 
inactivated more rapidly than the 15 minute complexes. 

Experiments with T7. A complete set of survival curves for monocomplexes 
of T7 and strain B irradiated at intervals during the latent period (12 minutes) 
is given in figure 7. It will be observed that the resistance remains unchanged 
(at the same value as for free phage) for 3 minutes in the presence of nutrient. 




200 
SECONDS 



300 



100 

UV DOSE IN 

Figure 7 . Survival curves for T7 complexes irradiated at different times during the 
latent period. Each curve is marked with the time in minutes after adding broth to the 
complexes formed in buffer. All data are from the same growth experiment at 37 C (latent 
period = 12 minutes). Average multiplicity of infection = 2 X 10~'. 



As time goes on, the curves become multiple-target in character, the average 
multiplicity continually increasing, while the final slope changes only slightly. 
This result stands in marked contrast to T2r and resembles that predicted by 
target theory. The exact shapes of these curves are not consistent with an as- 
sumption of equal multiplicities for all complexes at a particular time. However, 
the shape may be explained by a distribution of the number of targets, which 
may be expected as a consequence of different rates of phage multiplication in 
different cells. It is well known (Delbriick, 1945) that the number of phage 
particles released by lysis of individual cells has an extremely wide distribution. 



307 



1952] 



ULTRAVIOLET LIGHT INDEX IN BACTERIOPHAGE 



69 



To illustrate the effect of such a wide distribution upon our survival curves, a 
composite curve is plotted in figure 8, assuming a population consisting of equal 
numbers of cells having the various multiplicities in figure 1. By extrapolation 
to zero dose of the asymptotic slope of such a curve, it is possible, in principle, 
to determine the average multiplicity. However, this requires precise data at 
low survival values and cannot be done accurately with the points in figure 7. 




123456789 10 
DOSE 

Figure 8. Theoretical survival curves for complexes containing various numbers 

of targets (as in figure 1). 

Theoretical survival curve for a mixed population of complexes. The mixture is 

assumed to contain equal proportions of complexes having multiplicities 1, 2, 5, 10, 20, 50, 
and 100. 

Note: It is possible, by assumption of a suitable population containing a very wide dis- 
tribution of multiplicities, to obtain a composite theoretical survival curve which is expo- 
nential, simulating a "one-hit" curve (Dulbecco, personal communication). 

DISCUSSION 

The original intent of Luria and Latarjet's experiment, namely to observe 
the increase in the number of intracellular phage particles during the latent 
period, has been achieved with T7. The pattern of growth, so far as radiation 
resistance is concerned, does not exhibit the anomalies of (T2 and) T2r. There 
appears to be simply an increase with time of the average number of targets 
per cell, each target being similar radiologically to a T7 particle. 

This result shows that irradiation of intracellular phage, in the case of T7, 



308 



70 S. BENZER [vol. 63 

enables us to discern the existence of multiple intracellular entities at a time when 
fully infective phages are not yet detectable with the Doermann technique. 

The anomalous results with T2r may be of far greater interest, however. 
During the critical phase between the entrance of one infecting particle and the 
appearance of intracellular progeny, enormous changes are reflected by the 
survival curves. It seems likely that these resistance changes may bear a close 
relation to the phenomenon of multiplicity reactivation. While the resistance 
changes do not tell us what is actually happening, they at least give us something 
easy to measure which serves as an empirical index to development. The value 
of this index has already been demonstrated previously in the justification of 
the techniques used in these experiments (adsorption in buffer, chilling to stop 
growth). 

In the case of T2r, the maximum change in slope is by a factor of over 20 (free phage 
compared with t = 10 minutes). As pointed out by Luria and Latarjet, intracellular accumu- 
lation of ultraviolet-absorbing materials (nucleic acid components) must be considered as 
a possible contributing factor. 

By microspectrography of uninfected cells (at 2570 A) Hed^n (1951) finds an average 
extinction through the thickness of a cell of around 0.1, corresponding to 80 per cent trans- 
mission. A phage particle which is adsorbed on the surface of a bacterium is therefore 
shielded from ultraviolet to the extent of 20 per cent in one direction. This should lead to 
an average increase by only 10 per cent in the resistance (at i = 0) compared with free 
phage (since the cell rotates in all directions during exposure). Penetration of the phage 
into the cell should have little effect on the magnitude of this average shielding. In order 
to account for a subsequent change by a factor of 20 in the intensity of ultraviolet reaching 
the phage, a coating of nucleic acid, 1 ^ in diameter, would have to be produced. The meas- 
urements of Luria and Human (1950) and Cohen and Arbogast (1950) on optical density and 
nucleic acid content in suspensions of cells multiple infected with T2 reveal increases during 
growth, but these are far too small to account for the observed resistance changes during 
the first 10 minutes. 

In the case of T7, however, the small change in slope may well be due to a screening 
effect. 

A plausible interpretation of the increase in resistance of T2r complexes during 
the first half of the latent period may be the following: A T2r particle, after 
adsorption to a sensitive bacterium, must undergo a series of successive steps 
A^B— >C— ^D-^, etc. in the course of reproduction. Each of these steps has 
a certain cross section for being blocked by ultraviolet (e.g., by inactivation of 
an enzyme which is concerned with the step). Blockade of any one of these steps 
prevents normal development and causes inactivation of the phage. At time 
zero, the total cross section of the phage is therefore the sum of these individual 
cross sections, and the survival curve is exponential. As development proceeds, 
the steps which have already been passed are no longer needed, and the effective 
cross section decreases progressively while inactivation of the remaining steps 
retains the characteristics of a one-hit phenomenon. 

The absence of this behavior in the case of T7 suggests that there may be 
great differences in the mode of reproduction of T7 and T2r. These two phages 
are also dissimilar in other respects. T2r is a relatively large particle with a 
head and a tail and appears to have a kind of membrane (Anderson, 1949); it 



309 



1952] ULTRAVIOLET LIGHT INDEX IN BACTERIOPHAGE 71 

shows multiplicity reactivation with other particles of T2r, and also undergoes 
genetic recombination with mutants of T2 and other closely related phages. 
T7, on the other hand, is much smaller in size, spherical in shape, does not appear 
to have a membrane, and does not show multiplicity reactivation. Whether it 
undergoes genetic recombination is not known. 

Therefore, the sequence of events in the intracellular development of T2r, 
which is reflected in the resistance changes, may be characteristic for the mul- 
tiplication of only certain types of phage. 

We stand to learn a great deal about the growth of T2r by making use of the 
radiation resistance index. If conditions are suitable for phage development, the 
index (i.e., the survival of infective centers after a given standard dose of ultra- 
violet) goes up (or down, in the latter half of the latent period). If, for any reason, 
phage development is blocked, the index remains constant; its value marks the 
stage at wliich the block occurs. This makes it possible to study the effect, at 
any chosen part of the latent period (except near the end), of many factors, for 
example temperature, growth requirements, and specific chemical inhibitors. 
The results of these investigations will be reported separately. 

This tool offers promise not only for studying phage growth but for certain 
problems in bacterial physiology as well. The growth of phage in a bacterium 
is dependent upon the metabolic well-being of the cell and on its ability to make 
use of the substrates supplied to it. By infecting the bacterium with a T2r 
particle and using the radiation resistance index, the metabolic capabilities of 
the bacterium under particular conditions can be measured. A unique feature 
of this technique is that one can determine not only the average rate of metab- 
olism but the distribution of rates (from analysis of the survival curves) among 
the individual cells in the population. This idea is currently being applied to the 
problem of the kinetics of enzymatic adaptation in bacteria. 

SUMMARY 

Techniques are described for simultaneous starting and stopping of the growth 
of bacteriophage in all the host cells of a culture. Using these techniques, a 
comparison is made of the changes in resistance to ultraviolet of phages T2r 
and T7 during intracellular growth, following the method of Luria and Latarjet. 

Phage T7 gives results similar to expectations from target theory, while the 
results with T2r confirm the (anomalous) behavior observed by Luria and 
Latarjet, indicating that there may be large differences in the modes of repro- 
duction of different phages. 

The utility of the change in resistance as a tool in studying bacteriophage 
reproduction and certain problems in bacterial physiology is pointed out. 

REFERENCES 

Anderson, T. F. 1948 The growth of T2 virus on UV killed host cells. J. Bact., 56, 

403-410. 
Anderson, T. F. 1949 The reactions of bacterial viruses with their host cells. Botan. 

Revs., 15, 464-505. 



310 



72 s. BENZER [vol. 63 

CoHKN, S. S., AND Arbogast, R. 1950 The mutual reactivation of T2r+ virus inactivated 
by ultraviolet light and the synthesis of desoxyribose nucleic acid. J. Exptl. Med., 
91, 637-650. 

Delbruck, M. 1945 The burst size distribution in the growth of bacterial viruses. J. 
Bact., 50, 131-135. 

DoERMANN, A. H. 1948 Intracellular growth of bacteriophage. Carnegie Inst, of 
Wash. Year Book, 47, 176-182. 

DuLBBCCO, R. 1950 Experiments on photoreactivation of UV-inactivated bacteriophage. 
J. Bact., 59, 329-347. 

Hed^n, C. 1951 Studies of the infection of E. Colt B with the bacteriophage T2. Acta 
Path. Microbiol. Scand., Supplementum LXXXVIII. 

Latarjet, R. 1948 Intracellular growth of bacteriophage studied by Roentgen ir- 
radiation. J. Gen. Physiol., 31, 529-546. 

Luria, S. E. 1947 Reactivation of irradiated bacteriophage by transfer of self-repro- 
ducing units. Proc. Natl. Acad. Sci., 33, 253-264. 

Luria, S. E., and Dulbecco, R. 1949 Genetic recombinations leading to production of 
active bacteriophage from ultraviolet inactivated bacteriophage particles. Genetics, 
34, 93-125. 

Luria, S. E., and Human, M. L. 1950 Chromatin staining of bacteria during bacterio- 
phage infection. J. Bact., 59, 551-560. 

Luria, S. E., and Latarjet, R. 1947 Ultraviolet irradiation of bacteriophage during 
intracellular growth. J. Bact., 53, 149-163. 



311 



INVESTIGATIONS ON A LYSOGENIC BACILLUS 
MEGATERIUM 

hy 

Andre Lwoff and Antoinette Gutmann 

Department of Microbial Physiology of the Institut Pasteur 

I. Introduction 

We call lysogenic bacteria those bacteria which perpetuate the capacity to 
form bacteriophages without intervention of exogenous bacteriophages. This 
definition will be discussed and justified in the course of this work. Thanks to 
the investigations of J. Bordet, of F. Burnet and M. McKie, of Den Dooren de 
Jong, of E. and E. Wollman, and of J. Gratia, it is known: 1. that in a lysogenic 
strain lysogeny is an attribute of every bacterium and, in the case of sporogenic 
species, of every spore; 2. that lysogeny persists after repeated passages in the 
presence of a specific anti-bacteriophage serum; 3. that lysogenic bacteria adsorb 
the bacteriophages which they produce; 4. that lysozyme lysis does not liberate 
bacteriophages from potentially lysogenic bacteria; 5. that, starting from certain 
sensitive strains infected by a given bacteriophage, one can isolate lysogenic 
bacteria which produce a bacteriophage identical to the original bacteriophage. 

It thus seems strange that the very existence of lysogenic bacteria could have 
been put in doubt by A. D. Hershey and J. Bronfenbrenner (1948), who wrote: 
"How virus is transmitted from cell to cell in lysogenic cultures seemingly 
refractive to lysis remains to be clarified. It must be concluded, however, that 
the phenomenon of lysogenesis, frequently cited as evidence for the spontaneous 
intracellular origin of virus, can equally well be explained as some type of 
association between exogenous virus and incompletely susceptible bacterium." 
It seems that one must look for the origin of the skepticism in a possibly too 
narrow definition of "true" lysogeny due to M. Delbriick (1946), which restricts 
lysogeny to the case where the bacteriophage is liberated by bacteria in the 
absence of lysis. Since no example of this phenomenon is known, lysogeny 
vanishes. This "true" lysogeny was contrasted by M. Delbriick with "pseudo- 
lysogeny," in which bacteriophages are liberated following the lysis of sensitive 
bacteria, the latter being produced by mutation in the course of the growth of a 

Translated from the French and reprinted by permission of the 
authors and the Institut Pasteur from the Annales de l'Institut 
Pasteur, 78, 711-739 (1950). (N.B. A part of the section 
"Materials and Techniques" has been omitted from the translation 
of this paper.) 

312 



ANDRE LWOFF AND ANTOINETTE GUTMANN 

resistant culture contaminated by bacteriophages. But these arbitrary defini- 
tions do not exclude lysogeny in the sense in which we have defined it ; lysogeny 
in which the property to produce bacteriophage is hereditary. In 1925 J. Bordet 
wrote: "The faculty to reproduce bacteriophages is inscribed into the very 
hereditary weft of the microbe," and Burnet and McKie (1929b) expressed 
exactly the same idea: "The permanence of the lysogenic character makes it 
necessary to assume the presence of bacteriophage, or its anlage, in every cell of 
the culture, i.e., it is a part of the hereditary constitution of the strain." 

No doubt the experiments on which this conclusion was based — which shall 
be discussed in the course of this work — however suggestive they were, did not 
bring the element of certainty. It is true that as late as 1949, in spite of numerous 
works published on lysogenic bacteria, one still did not know the manner in which 
lysogenic bacteria liberate the bacteriophage that they produce. Is the bacterio- 
phage secreted by the bacteria in a continuous manner? Is it secreted at the 
moment of division? Is it liberated by the lysis of certain bacteria? In the 
absence of any facts, the majority of microbiologists envisaged, along with 
Bordet and Renaux (1928), and with Burnet and McKie (1929b), that the 
bacteriophage is "secreted" by the bacterium. But it is really of little importance 
for the definition of lysogeny whether the bacteriophage is liberated in this or any 
other manner. The important fact is the reproduction of a bacterium potentially 
capable of producing a bacteriophage, that is to say the hereditary capacity to 
produce a bacteriophage. Naturally this does not exclude the mode of liberation 
of the bacteriophage from being worthy of our attention. 

J. Northrop (1939) established a parallelism between enzymes and bacterio- 
phages, based on the observation that the concentration of bacteriophages in a 
culture of lysogenic bacteria increases, like that of the enzyme gelatinase, in an 
essentially parallel manner to the multiplication of the bacteria. But it is quite 
evident that the interpretation which one can draw from such facts will be very 
different according to whether, for example, in the interval of two divisions every 
one of 100 individuals of a bacterial population liberates a bacteriophage, or 
whether a single bacterium liberates 100 bacteriophages. 

The principal problems posed by the existence of lysogenic strains are the 
following: 1. Can the faculty of producing bacteriophages really be perpetuated 
without intervention of exogenous bacteriophages? 2. How do lysogenic bacteria 
liberate the bacteriophages which they produce? 3. If the production of bacterio- 
phages appertains to only a certain proportion of the bacteria, then what factors 
induce the production of bacteriophages in a population of potentially lysogenic 
bacteria? The experiments presented here in detail were summarized in four 
preliminary notes (Lwoff and Gutmann, 1949-1950) ; they furnish a reply to the 
first two questions. The third question was studied in collaboration with 
L. Siminovitch and N. Kjeldgaard, and the relevant results shall be the subject 
of a separate paper. 

II. Materials and Techniques 

It was evident a priori that the study of "mass cultures," as it is usually 



313 



INVESTIGATIONS ON A LYSOGENIC BACILLUS MEGATERIUM 

practiced, can only furnish partial solutions to the problems of lysogenic bacteria. 
Only observations carried out on individual bacteria, or on microcultures 
containing a small number of individuals, are capable of leading to definite 
conclusions. 

Strain. Our experiments were carried out on a lysogenic strain of Bacillus 
megaterium. We have used the classic strain "899" of den Dooren and de Jong 
and an asporogenic sensitive mutilat strain. Lysogenic and mutilat strains were 
subcultured by passages on peptone broth agar. 

Technique of Micromanipulation. One utilizes an incubator box made from 
transparent plastic material. An electric heater is placed under the bottom of 
this box, into which 20 or so holes have been drilled. The temperature is thus the 
same in all parts of the incubator box and a thermoregulator in one corner insures 
a constant temperature of 37°. The microscope and the receptor for the de 
Fonbrune micromanipulator are fixed on the base of the Zeiss-Peterfi micro- 
manipulator. We have utilized the oil chamber of de Fonbrune. It is unnecessary 
to go into details concerning this device, of which one can find a description in the 
monograph of de Fonbrune (1949). A knowledge of these technical details, 
however, can prevent considerable loss of time. 

Two pieces of glass of 1.5 X 12 X 7 mm. are glued to either side of the depres- 
sion slide. Their separation is so calculated that it exceeds the length of the 
coverslips by about .8 mm. The latter, made from neutral glass, are washed in 
nitric acid, rinsed in double distilled water, wiped with fine cloth and sterilized 
(as well as the slide) at 150°. 

For the preparation of the experiment: 1. one sets a cover slip on the slide and 
places several drops of paraffin oil on the cover slip, so that it is covered by a layer 
of oil. 2. One draws out the fine end of a Pyrex glass Pasteur pipette so that the 
diameter of the terminal capillary allows passage of culture medium and that 
when the tip of the pipette filled with liquid comes in contact with the cover slip 
under the oil, one drop of liquid comes out and spreads out on the glass. This 
drop must have a diameter of about 500 to 800 m- Nine rows of 9 drops can thus 
be regularly placed. The second and third rows are interrupted at the point of 
the fourth, fifth, and sixth drop. It is of course possible to place several different 
media on the cover slip. The cover slip is then turned upside down and oil is 
poured between it and the slide. On the bottom of the oil chamber one places 
drops of distilled water in a manner so as not to disturb the observations; this 
will prevent any significant evaporation of drops of less than 100 m diameter. 
The oil chamber is then placed on the stage of the microscope. 

Technique of Sampling. It is necessary that one can transfer one bacterium, or 
a minimum quantity of liquid of the order of 10"^ mm.^ into a tube or onto a 
petri dish. The dimensions of the liquid column sampled can be measured with 
the micrometer objective. In order to transfer the content of a pipette into the 
tube, one can proceed in the following manner : 1 . The field of the microscope is 
disengaged by placing the oil chamber towards the rear. 2. The needle and the 
needle carrier are turned by 90°, in such a manner that the point of the needle 
points towards the right. The angle formed by the micropipette is then in a plane 



314 



ANDRE LWOFF AND ANTOINETTE GUTMANN 

parallel to that of the stage of the microscope. The pipette is then displaced 
towards the left, out of the microscope field. 3. A Pasteur pipette whose tip has 
been bent at a right angle about two to three mm. from the end is introduced into 
the right needle carrier of the Peterfi micromanipulator. The pipette contains 
dilution medium for a length of about 2 cm. It is so oriented that the angle 
formed by the terminal bend is in a plane parallel to the plane of the stage. 
4. The tip of this pipette is brought into the center of the microscope field. The 
micropipette is then moved towards the right and its level regulated in such a 
manner that its point is opposite the liquid meniscus of the Pasteur pipette. One 
introduces the micropipette into the liquid. One inoculates the content of the 
micropipette into the liquid column of the Pasteur pipette. If one manipulates 
the bacterium, it is possible to see that bacterium pass from micropipette to 
Pasteur pipette. In any case, one knows that the operation is finished when a 
drop of oil appears at the point of the micropipette. All these manipulations are 
carried out in the incubator box. 5. The Pasteur pipette is then detached from 
its support, removed from the box and its contents transferred either into a tube 
or onto a petri plate. After a little practice, these operations can be performed 
very quickly and with great assurance. All of these manipulations are naturally 
carried out aseptically. According to our experience, which includes almost 1000 
samplings, the chances of contamination are practically nil, since one manipulates 
liquid quantities of the order of 10~^ to 10~^ mm^. 

III. Bacterial Multiplication 
WITHOUT Liberation of Bacteriophage 

a. A diplo-bacillus, after five washings, is inoculated into a drop of peptone 
water. After each division, one of the daughter cells is removed with a minimum 
of the liquid. Twenty-two daughter cells are thus successively removed. The 
entire drop is then removed: no bacteriophages. 

b. In another experiment of the same type, one removes successively 42 
daughter cells: the result is identical. Furthermore, assays of four samplings in 
the course of the experiment and of the total sample at the end show no bacterio- 
phages. 

c. The washed filament of three bacteria is observed. One removes successively 
14 daughter cells and makes 5 samplings. After the last division, the entire fluid 
is assayed: no bacteriophages. 

d. Three diplo-bacilli are washed and inoculated separately into drops of 
peptone water. When the number of bacteria has attained respectively 24, 47, 
and 54, the entire fluid is sampled. There are no bacteriophages in the fluid nor 
in any of the samples previously taken. 

e. Three diplo-bacilli are washed and transferred into drops of peptone water 
of 72 to 100 mil diameter. Multiplication stops spontaneously when the number 
of bacteria reaches respectively 8, 12 and 16. All of the fluid is removed and 
sampled: no bacteriophages. 

f. A filament of 7 bacteria is washed and inoculated into a synthetic medium. 



315 



INVESTIGATIONS ON A LYSOGENIC BACILLUS MEGATERIUM 

After two hours, there are 82 bacteria: all of the liquid is sampled, no bacterio- 
phages. 

Discussion. These experiments, as well as those which shall be described in the 
two following paragraphs, lead to the thought that a lysogenic bacterium 
multiplies without producing bacteriophages. They demonstrate that B. 
megaterium 899 can grow and multiply without liberating bacteriophages. Thus, 
liberation of bacteriophages is not a phenomenon which necessarily accompanies 
growth of a population of a lysogenic bacteria ; it appears, rather, to be a contin- 
gent phenomenon. 

IV. Maintenance of Lysogenic Power 
IN THE Absence of Free Bacteriophage 

A motile diplo-bacillus is sampled from a culture in peptone water (proteose 
peptone at 1%). This bacillus is successively transferred in six drops of the same 
medium. It is left in the sixth. During the entire length of this experiment the 
diplo-bacillus has remained motile. With B. megaterium the separation of the 
daughter members generally takes place when the daughter bacteria are already 
partitioned off. Once this separation releases two daughter diplo-bacilli, one 
injects into the drop approximately \ of its volume of new medium. One assures 
a mixing of the fluids by repeated aspiration back and forth of approximately 
half of the fluid and then removes one of the bacteria in a volume essentially equal 
to the volume of fluid which has been added. This sample is transferred into a 
suspension of sensitive mutilat in soft agar and this suspension is then spread on 
an agar surface. This operation was repeated until the 10th division. At this 
point, the two daughter bacteria were sampled separately, the latter with the 
total of the residual fluid. The 11 plates that had been inoculated all showed a 
single plaque centered by a bacterial colony, with the exception of the first and 
the eighth, which showed no plaque whatsoever. The absence of the lysogenic 
colony on the two negative "plates" in this experiment, as in the following, 
probably results from an accidental loss of the bacterium. This experiment was 
repeated in peptone water to which glucose and a liver extract had been added. 
The original diplo-bacillus was followed until the 19th division, after which the 
two final bacteria were plated. The 20 plates, with the exception of the 4th and 
the 16th which showed no plaque whatsoever, each had, as in the previous series, 
one plaque centered by a lysogenic colony. One of these centered colonies, after 
being picked and isolated, proved itself to be lysogenic: the filtrate of its culture 
contained bacteriophages. 

The absence of any plaque, except for the single plaque centered by a bacterial 
colony, shows that no free bacteriophages ever appeared in these microcultures. 
During the experiments, there was never any noticeable decrease in the size of 
the diplo-bacillus, as would have occurred if one of its members had lysed, nor 
was lysis ever observed. Incidentally any lysis would have brought some 
disturbance in the rhythm of the divisions; but the 19 divisions took place in 
486 minutes at regular intervals of 25 ± 3 minutes. 

Discussion. It was established by Bordet and Renaux (1928) that lysogenic 



316 



ANDRE LWOFF AND ANTOINETTE GUTMANN 

strains of E. coli conserve their lysogenic nature after numerous passages in the 
presence of an antibacteriophage serum. Analogous results had been obtained 
by F. Burnet and M. McKie for various bacteria (1929a,b) and by E. and E. 
Wollman (1936) for B. megaterium. Since the antibacteriophage serum neutral- 
izes free bacteriophages, these experiments demonstrate that lysogeny can 
maintain itself in the absence of free bacteriophages, provided that the concen- 
tration of antibacteriophage serum during the entire length of the experiment is 
sufficient to immediately inactivate all bacteriophages. In other words, one must 
be sure that, under the conditions of these experiments, there is no possibility of 
a free bacteriophage infecting a bacterium before being neutralized by the serum. 
This condition was probably met in some of these experiments, but this is not 
a priori certain. And nothing enables us to state categorically that bacterio- 
phages which adsorb to bacteria are actually inactivated by serum. 

It is also known that lysogeny maintains itself equally well in cultures growing 
in media supposedly deprived of calcium; in fact, in media containing oxylate or 
citrate (E. and E. Wollman, A. Gratia). But numerous organic substances 
possess the power to form complexes with calcium; these complexes could then 
dissociate, and, in organic media, it is difficult to ascertain the absence of free 
Ca"''"'' ions. The fact that lysogeny of a culture maintains itself in a citrate 
medium, whereas the adsorption of bacteriophage does not take place except in 
the presence of calcium, is but an argument in favor of its endomicrobial perpet- 
uation. 

Finally, it was shown by Den Dooren de Jong that the lysogenic function of 
B. megaterium is transmitted in cultures which have been derived from a spore, 
and even from a heated spore. This is true for all sporogenic lysogenic strains. 
This is, thus, an additional argument in favor of the endomicrobial transmission 
of the bacteriophage. But, unfortunately, the form of the bacteriophage within 
the interior of the spore is not known. If one adds to all of these considerations 
the fact that in a lysogenic culture every colony — supposedly obtained from a 
single cell — is lysogenic, one sees that a considerable accumulation of facts 
favors the conception that the faculty to form bacteriophages is hereditary, that 
is to say, independent of any exogenous reinfection. In our experiments, lysogeny 
is maintained in the course of 19 divisions in the absence of free bacteriophage. 
If the bacteriophage does not multiply within the bacterium and if the mainte- 
nance of lysogeny proceeds from particles adsorbed on the original diplo-bacillus 
and distributed equally over the daughter bacteria, then that diplo-bacillus must 
have carried a minimum of 2^^ or 524,288 bacteriophages at the onset. The 
chance of an unequal repartition over the daughter cells in combination with our 
selection of that bacterium which happened to have inherited the smallest 
number of bacteriophages after each such division is so small that it need not be 
taken into consideration. 

S. Bayne-Jones and L. A. Sandholzer (1933) estimated the average volume of a 
single B. megaterium as 2.5 n^. Let us assume the value of 6 m^ for the average 
diplo-bacillus. According to McLauchlan, E. M. Clark, and F. W. Boswell 
(1947), the dimensions of the head of the bacteriophage of Bacillus megaterium 



317 



INVESTIGATIONS ON A LYSOGENIC BACILLUS MEGATERIUM 

are 40 by 100 van, that is to say, a volume of approximately .9 X 10"^ ii^. Let us 
suppose that the polyhedral bacteriophages completely fill the entire diplo- 
bacillus, which could then contain no more than 66,000 bacteriophages. But, in 
any case, the maximum number of bacteriophages produced by any single 
bacterium never seems to surpass 1000. 

One can consider, without much chance of error, that in a viable lysogenic 
bacterium the space occupied by the bacteriophages must not surpass 50% of 
the total bacterial volume. An average viable lysogenic B. megaterium diplo- 
bacillus could thus contain maximally 33,000 bacteriophages. This figure is 
significantly lower than the value of 524,288 which represents the minimum 
number of particles which must have been present in the original diplo-bacillus 
in order to sustain the hypothesis of the perpetuation of the lysogenic function by 
a simple distribution of pre-existing exogenous particles. This hypothesis thus 
can be rejected. 

The hypothesis, according to which lysogeny perpetuates itself by an endo- 
microbial route without the necessary intervention of exogenous bacteriophages, 
seems to us, for the moment, the only one capable of accounting satisfactorily for 
the experimental facts. The problem of the form in which the lysogeny perpet- 
uates itself shall be discussed in the course of this work. Another conclusion can 
be drawn from our experiments. One knows (J. Bordet and J. Renaux, 1928; 
F. Burnet and M. McKie, 1929a,b) that all colonies issuing from a lysogenic 
culture are lysogenic. Our experiments confirm and validate these data: every 
diplo-bacillus from a population of B. megaterium 899 isolated by means of the 
macromanipulator gives rise to a lysogenic colony. 

V. Lysozyme Lysis Does Not Liberate Bacteriophage 

F. Burnet and M. McKie (1929a), F. Burnet and D. Lush (1936) observed that 
lysis by a bacteriophage X of a lysogenic bacterium sensitive to this bacterio- 
phage X and normally producing a bacteriophage Y does not liberate the 
bacteriophage Y. F. Burnet and M. McKie understood the whole significance of 
this phenomenon as early as 1929 and concluded that all bacteria of a perma- 
nently lysogenic strain possess in their hereditary constitution some unit which is 
potentially capable of liberating phage. Burnet and McKie found, nevertheless, 
that a small proportion (about 1%) of the bacteria, of a 24-hour-old agar culture 
harbors bacteriophages which can be liberated by treatment with distilled water. 
E. and E. Wollman and A. Gratia (1936), independently and within a few weeks of 
each other, made a similar observation: lysozyme lysis of lysogenic B. megaterium 
899 does not liberate bacteriophages. E. and E. Wollman found one bacterio- 
phage for several thousand bacterial colonies and concluded that the bacterio- 
phage exists in two different forms, phases, or states: the "mature" bacterio- 
phage which one could call corpuscular, or, more simply, bacteriophage-virus as 
it appears in the free state in the culture medium, and the non-virulent, latent 
"intracellular bacteriophage" as it exists inside lysogenic bacteria and which 
we call probacteriophage. Furthermore, it is known that it is impossible to 
liberate by lysozyme lysis bacteriophages adsorbed to living B. megaterium cells. 



318 



ANDRE LWOFF AND ANTOINETTE GUTMANN 

In contrast, lysozyme lysis does liberate bacteriophages adsorbed to killed 
bacteria (A. Pirie, 1940). The bacteriophage virus does not transform itself into 
probacteriophage except in living bacteria. 

The importance of these facts stimulated us to verify the action of lysozyme 
lysis. Washed bacteria are placed into microdrops. In the course of exponential 
multiplication, one injects into the microcultures a solution of lysozyme of such 
concentration that its final dilution is 1 to 50,000. The bacteria are lysed in 
several minutes. In different experiments 12, 34, 80, and 208 bacteria were thus 
lysed. All of the liquid was sampled: no bacteriophages. The presence of 
lysozyme at the concentration of 1 to 50,000 in a microdrop does not interfere 
with the assay because the medium is considerably diluted. These observations 
thus extend to exponentially growing bacteria the results of E. and E. Wollman 
and of A. Gratia on B. megaterium and they confirm the fundamental observation 
of F. Burnet and M. McKie (1929b). 

VI. How Do Lysogenic Bacteria Liberate Bacteriophages? 

We are thus in possession of the following facts: 1. In the lysogenic strain of 
B. megaterium 899, every bacterium isolated in the micromanipulator and 
washed gives rise to a lysogenic colony. 2. The faculty of producing bacterio- 
phages perpetuates itself in the absence of exogenous bacteriophages, that is to 
say by an endomicrobial route. 3. Bacteria can multiply without liberating 
bacteriophages. 4. Lysozyme lysis does not liberate bacteriophages. 

Nevertheless, a filtrate of a culture of lysogenic B. megaterium, indeed the 
filtrate of every lysogenic culture, always contains some bacteriophages. We 
have verified the fact that this is also true for dense cultures in microdrops: one 
always finds bacteriophages in the liquid of a small drop of peptone water which 
contains several thousand bacteria. One can thus entertain the hypothesis that 
only a small percentage of the bacteria produce bacteriophages. We were 
encouraged to persevere in the investigation of the mode of liberation of bacterio- 
phages by the observation of bacterial "ghosts." The formation of the ghosts 
was observed in the microdrops. It is the result of a relatively slow degeneration 
of the bacterium which can take several minutes or several seconds. The ghosts 
persist for several hours. We thought that we had observed bacteriophage lysis. 
But assay of the medium after the slow lysis did not reveal any bacteriophages. 
Nevertheless, we continued to follow the development of the microclones, 
assaying the medium from time to time. We give here the results of a first series 
of experiments published previously in a preliminary note (A. Lwoff and A. 
Gutmann, 1949b). 

A. A filament of free bacteria is washed and brought into one drop of medium 
S. The following table gives the time in minutes (m), and the number of bacteria 
(b): 

m h m h 






3 


30 


4 


55 


6 



65 


7 


105 


25 


123 


45 



319 



INVESTIGATIONS ON A LYSOGENIC BACILLUS MEGATERIUM 

After the 123rd minute one removes 28 X 10'' /x^ of medium, representing 
approximately 90% of the volume of the drop: 480 bacteriophages. The 
bacteria remained united in the long filament. No lysis had been observed. 
No ghost was visible. We thus concluded that it might be possible that the 
production of bacteriophages proceeds in the absence of bacterial lysis. At this 
time we knew only one type of lysis: the slow lysis in which a well recognizable 
ghost persists and we thought that the absence of such a ghost implies the absence 
of lysis. Since then, we have recognized another rapid type of lysis, in the course 
of which the bacteria disappear in less than one second without leaving a recog- 
nizable residue. This is the type of lysis which liberates bacteriophages. Several 
terminal bacteria could well have lysed in the course of this experiment, and this 
lysis could have gone unnoticed. This experiment can thus not be retained to 
support the thesis of a liberation of bacteriophage without bacterial lysis. 

B. A diplo-bacillus is washed and placed in peptone water. One notices, 
according to the table below, that one bacteria disappeared between the 50th 
and the 55th minute, two disappeared between the 61st and the 62nd minute: 



m 


h 





2 


4 


4 


14 


4 


31 


6 


35 


7 


50 


11 


55 


10 


61 


13 


62 


11 


83 


17 


109 


43 



Samples were taken from the microdrops after 14, 43, and 55 minutes respec- 
tively 2.5 X 10^ 4 X 10^ and 2.4 X 10^^: no bacteriophages. At the 63rd 
minute, one samples 2.4 X 10^ /x^: 18 bacteriophages. At the 83rd minute, 
3.2X10^^:17 bacteriophages. At the 115th minute 5.15X10^^:68 
bacteriophages. The residual volume was 92 X 10^ /x^- The initial volume was 
thus 149 X 10^ M^- One can thus calculate that the number of bacteriophages at 
the time of the three samplings was 11 30, 725, and 190. It is to be noted that the 
sampling made at the 55th minute, after the disappearance of one bacterium, did 
not reveal any bacteriophages. In this experiment the bacteria were motile. It is 
possible that the first lysis was a slow lysis which did not produce bacteriophages 
and in which the bacterial ghost remained unnoticed. One also notices that the 
calculated number of bacteriophages was 1,130, a number which seems to us too 



320 



ANDRE LWOFF AND ANTOINETTE GUTMANN 

high to correspond to the lysis of only two bacteria. It is possible that, in this 
experiment, the mixing before the first sampling was insufficient. 

C. A filament of three bacteria is washed and transferred into peptone water. 
The history of these three bacteria, a, h, and c, which remained in a single 
filament, is indicated in the table below. One notices that bacterium a divided 
twice, and that the four bacteria which issued from its division lysed between 
the 88th and the 114th minute. 



m 


a 


b 


c 





1 




1 


25 


1 




2 


55 


2 




2 


77 


2 




2 


82 


4 




3 


88 


4 




4 


107 


2 




4 


114 





2 


4 


120 




2 


6 


134 




2 


9 


142 




2 


9 


155 




2 


11 



A sampling of 1 X 10^ /x^ undertaken at the 114th minute did not reveal 
bacteriophages. A sampling of 1.4 X 10^ at the 142nd minute revealed the 
presence of 71 bacteriophages. Sampling the total fluid at the 155th minute 
(116.3 X 10* n^) revealed 400 bacteriophages. If one takes into account the ratio 
of these volumes, one concludes that the total number of bacteriophages must 
have been approximately 900 at the instant of the 2nd sampling. It is to be noted 
that the first sampling at the 114th minute was negative while at the 142nd 
minute there was the appearance of numerous bacteriophages without there 
having been any lysis of other bacteria. It is possible that also here, as in the 
preceding experiment, the mixing of the culture medium was insufficient. It is 
also possible that after lysis the bacteriophages remained engulfed in the viscous 
cellular debris for some minutes and thus escaped the action of mixing currents. 

Another experiment should allow us to recognize the character of the lysis by 
the bacteriophage. A diplo-bacillus is washed and inoculated in one drop of 
proteose peptone. At the 26th minute, there are two diplo-bacilli, of which one 
is removed. At the 72nd minute, two other diplo-bacilli are removed. There a 
filament containing 4 bacteria remains in the drop. A sampling of 6.5 X 10^ ju^ 
made at this moment is negative; at the 143rd minute there is a filament of 8 and 
a filament of 10 bacteria. One sees one of the bacteria — perhaps one diplo- 
bacillus? — • disappear in about 1 second without leaving any trace. One samples 



321 



INVESTIGATIONS ON A LYSOGENIC BACILLUS MEGATERIUM 

8.5 X 10^ M^ of fluid, corresponding to about half the volume of the drop: 110 
bacteriophages. The total number of bacteriophages produced is thus about 220. 
It is this observation which convinced us of the existence of a rapid lysis followed 
by the appearance of bacteriophages. Subsequently, we have often observed 
lysis followed by the appearance of bacteriophages. We have noticed that this 
lysis does not leave any visible ghost and that it often occurs in less than one 
second. It could thus very well pass unobserved in a drop which contained 
several diplo-bacilli in the process of multiplication. 

We had thought (1949b) that we observed the liberation of bacteriophages 
without lysis, but after having recognized the two types of lysis — slow and rapid 
— and having analyzed the possible causes of error which can arise in the course 
of experiments in microdrops, we have attempted experiments in which a very 
small number of bacteria shall be present at the same time in a single microdrop. 
Each time that this was possible, we have worked in such a manner that only a 
single diplo-bacillus was present in a drop. It is only under these conditions that 
one can be assured that bacterial lysis is not overlooked. 

We have also tried to achieve conditions under which a single diplo-bacillus, or 
a single bacterium, lyses in one drop. In this way, one can sample the totality of 
the microdrop, thus avoiding having to measure the dimensions of the micro- 
pipette and calculating its volume and reducing the causes of error in the quanti- 
tative determinations to a minimum. We took care to stir the drop repeatedly if 
only a part of the drop was to be sampled. One will see in the following that no 
liberation of bacteriophage was observed in the absence of bacterial lysis. 

Second Series of Experiments, a. A filament of 14 bacteria is washed and 
transferred into a microdrop of peptone water at time h; at /i + 20 minutes, there 
are 18 bacteria. A diplo-bacillus detaches itself and is transferred into a micro- 
drop. Its two members lyse at /i + 40 minutes, at 40-second intervals. At 
h -\- QS minutes, 8 bacteria of the original drop are transferred into microdrops, 
where they lyse without leaving a trace in less than 10 minutes. There remained 
8 bacteria in the original drop. Seven of them lysed under our eyes between 
/i + 67 and h -\- 71. The residual bacterium is eliminated. All of the liquid is 
then sampled. The plaques on the petri dish are almost confluent, impossible to 
count. It looked as if there were about 2000 plaques on the plate. 

b. A filament of three bacteria is washed and brought into a drop A. At time 
h + 45, there are about 4 bacteria. A sampling made then is negative: no 
bacteriophages. At /i -f 60 minutes, there are 6 bacteria, only 4 at h -{■ 72. 
These four residual bacteria are transferred with a minimum of fluid into one 
drop B. All of drop A is sampled: 138 phages, corresponding to the lysis of two 
bacteria. In drop B two out of four bacteria lysed between h -\- 72 and h -f- 79. 
The two residual bacteria are transferred with a minimum of fluid into a drop C. 
All of drop B is sampled, where two bacteria have lysed: 38 bacteriophages. 
The two bacteria remain for 28 minutes in drop C. They are then transferred 
into D. The total sampling of C does not reveal any bacteriophages. In drop D 
the two bacteria disappear 4 minutes after their transfer: total sampling indicates 
162 bacteriophages. One will notice that in drop C, where two bacteria remained 



322 



I 



ANDRE LWOFF AND ANTOINETTE GUT MANN 

until the fourth minute preceding lysis, there was no liberation of bacteriophages. 

c. A diplo-bacillus was washed and brought into a drop A. Samplings made 
at h + 35 and /i + 70 minutes are negative. At /i + 120 minutes there are four 
bacteria. At the 140th minute, lysis of two members of a diplo-bacillus is 
observed within a 10-second interval. The other two bacteria are transferred 
into drops B and C. Total sampling of A: 27 bacteriophages. The bacterium in 
drop B disappears after the 12th minute: 9 bacteriophages. After 17 minutes in 
drop C, the remaining bacterium is transferred to D, from where, after an 
additional 17 minutes, it is transferred into drop E. All of drops C and D are 
sampled: no bacteriophages. In drop E, the bacterium has disappeared after the 
third minute: 15 bacteriophages. During the 34 minutes that had preceded lysis 
there was thus no liberation of bacteriophages by the bacterium which had 
liberated bacteriophages in drop E. 

d. A diplo-bacillus is washed and put under observation in a microdrop. At 
/i -|- 3 minutes there are three bacteria, at A -|- 48 minutes: 7 bacteria. At the 
65th minute, each of the bacteria has divded in two. There are eight bacteria 
which have remained together in one filament. At /i + 83, one of the bacteria is 
lysed. Within 6 minutes, three other bacteria have lysed. The four bacteria 
which have disappeared belong to the same mother bacterium . All of the fluid is 
sampled: 579 bacteriophages. The four remaining bacteria were transferred into 
another drop: after 22 minutes there were 10 bacteria whose fate was not 
followed further. 

Lyses were observed in numerous other experiments, whose details are unneces- 
sary to give here. The number of phages liberated by the lysis of a single bac- 
terium can be inferred from the following data, where the italicized numbers 
represent the number of bacteria lysed and the romanized numbers the number of 
bacteriophages liberated: 2: 138, 2: 38, 2: 162, 2:27 , 1 ■.^, 1: \b, 4- 89, 4: 452, 
.5:235, 1: 158, 2: 108, 1 : 178, 4:579, ^: 600. 4: 166; this corresponds to an 
average of 72 phages per bacterium lysed. The lowest number was 9, the highest 
178. It should be noted that under our observational conditions of utilizing an 
objective ocular lens of magnification 40 and 12.5 respectively, the septum which 
separates the two daughter bacteria is not visible at the time when it is already 
functional. 

We often happened to observe the lysis of a bacterium of unique appearance. 
Half of that bacterium, in fact a diplo-bacillus, disappeared under our eyes, 
leaving for a short time a mutilated bacterium bounded by a circular surface. 
Bacteriophage lysis thus reveals the existence of the septum although the latter 
is not yet visible under the conditions of our observations. It is quite exceptional 
to see single independent bacteria in the case of B. megaterium. Nevertheless, the 
separation of the two members of a diplo-bacillus was frequently observed before 
the lysis by the bacteriophage. We thus believe that certain of the numbers 
which we have given really correspond to the production of phage by a single 
bacterium. It remains possible, however, that other values which we have 
reported for one bacterium really correspond to a diplo-bacillus. 

An important statement must be made at this time. We have seen under our 



323 



INVESTIGATIONS ON A LYSOGENIC BACILLUS MEGATERIUM 

eyes some hundreds of bacteriophage lyses. We have thus observed the lysis of 
16, 8, 4, or 2 contiguous bacteria. But we have never observed the lysis of a 
single member of the diplo-bacillus. This question shall be discussed in paragraph 
VII. 

Discussion. J. Bordet and Renaux (1928), F. Burnet and M. McKie (1929b), 
and E. WoUman (1936) envisaged that the bacteriophage is secreted by lysogenic 
bacteria without the latter being destroyed. The problem of the mode of 
liberation of bacteriophage by lysogenic bacteria was also discussed by J. 
Northrop (1939) who, having noticed the appearance of bacteriophages during 
the growth of lysogenic cultures, concluded that the bacteriophage "is produced 
during the growth of bacteria and not during the lysis." M. Delbriick and 
S. E. Luria (1942), on the basis of the observations of E. Cordts, which, to our 
knowledge, have never been published, thought that lysogenic bacteria can 
produce bacteriophages without lysis. 

It should be noted that in the experiments of J. Northrop the value of the ratio 
of bacteriophages to bacteria was approximately 2. If, in the interval between 
two divisions, three bacteria out of every hundred lyse, each liberating 72 
bacteriophages, then the value of the ratio of bacteriophages to bacteria will be 
216 to 97; that is to say, approximately 2. Lysis of 3% of the bacteria during an 
interval of two divisions does suffice to account for the relative number of 
bacteriophages and bacteria observed by Northrop. This lysis would entail only 
a 3% reduction in the theoretical "growth rate." But it is difficult to measure 
growth rate with greater accuracy than 5%. Unless a very high proportion of the 
'bacteria lyses, no study of growth curves can thus furnish valid data relevant to 
the mode of liberation of the bacteriophages. All that can be said is that, in the 
experiments of Northrop, the bacteriophages are liberated during the growth of a 
bacterial population. But it is impossible to rule out the lysis of a certain percent- 
age of the bacteria. 

Our experiments thus bring proof that the bacteriophage produced by the 
lysogenic B. megaterium is liberated by bacterial lysis. Until the third minute 
prior to lysis no bacteriophage is liberated. For a long time bacteriologists have 
observed the partial lysis of colonies of lysogenic bacteria. The majority of these 
workers thought that this lysis was due to the bacteriophage. But we know that 
lysogenic bacteria can lyse without necessarily producing bacteriophage. Cultures 
of B. megaterium 899 lyse completely in peptone medium if they are deprived of 
oxygen. This lysis begins as soon as anaerobiosis is established. The lysis does not 
liberate bacteriophages. The fact that the colonies of a lysogenic bacteria lyse 
partially or totally does not imply that this lysis corresponds to the production of 
bacteriophages. A fortiori, it is not possible to conclude that under these condi- 
tions the bacteriophage is liberated by the lysis of bacteria. 

Is it possible to exclude the hypothesis of a secretion of bacteriophages by 
viable bacteria from the fact that in our experiments liberation of bacteriophages 
took place only by bacterial lysis? In the case of sensitive Escherichia coli 
infected by bacteriophage T2, the nucleus disappears between the fifth and tenth 
minute after infection (Luria and Palmer, 1945-1946). If the disintegration of 



324 



ANDRE LWOFF AND ANTOINETTE GUTMANN 

the nucleus obligatorily accompanies the production of bacteriophages, then it is 
evident that no bacterium could survive phage production. But it is possible to 
conceive of production of bacteriophages without destruction of an essential 
bacterial organelle. It is necessary, however, to think of a way for the bacterio- 
phages to come out. One knows that lysogenic bacteria adsorb the bacterio- 
phages. One is thus obliged to envisage a variation of bacterial properties which 
allows certain bacteria to adsorb bacteriophages at certain stages and to let them 
out at other stages. It is also possible to conceive that the bacteriophage is 
liberated when all the elements capable of fixing the bacteriophage are "satu- 
rated." Thus liberation of bacteriophages in the absence of bacterial lysis is 
theoretically possible, but at the present time we do not know of any such case.^ 
And in the only case in which the mode of liberation of bacteriophages by a 
lysogenic strain was studied with techniques offering sufficient guarantees, the 
mode of liberation of bacteriophages was bacterial lysis. 

VII. Factors of Bacteriophage Production 

The reader has no doubt noticed that in certain of our experiments all of the 
bacteria put into a little drop grew and multiplied, whereas in others all the 
descendants of one bacteria lysed. We thought at the outset that the sudden 
change of medium was responsible for the onset of bacteriophage production. 
In order to explain the fact that only a certain percentage of the bacteria was 
lysed, the hypothesis was envisaged according to which this change only induced 
the lysis if it took place at certain stages of the life cycle of the bacterium; for 
instance, during nuclear division. But, we subsequently observed considerable 
differences between different series of the experiments. On a certain day, for 
example, we had put 18 bacteria in three drops; all of them divided. The 153 
bacteria from these divisions were lysed by lysozyme: no bacteriophages. There 
had been no production of bacteriophage. On another day, we started with 38 
bacteria divided among 5 drops. The descendants of three groups did not show 
any lysis. The descendants of two groups did lyse partially, and all of the 
descendants of two other groups lysed. The following results correspond to the 
study of 104 groups of bacteria: 

Initial number 6 4 8 4 8 4 4 

Maximum number 17 41 29 13 34 8 10 

Number of lysed bacteria 12 14 8 10 

Proportion of lysed bacteria.... 0/17 0/41 0/29 12/13 14/34 8/8 10/10 



iW. H. Price noticed a massive liberation of bacteriophages by cultures of Staphylococcus 
muscae of constant optical density. The ratio of bacteriophages to bacteria rises from 0.2 
to 26 in about 30 minutes. A similar phenomenon was observed in cultures of Bacillus rnega- 
terium by A. Lwoff, L. Siminovitch, and N. Kjeldgaard (unpublished observations), who 
interpret the phenomenon in the following manner: The optical density of the culture remams 
fairly constant, or increases slightly, but the bacteria lyse in the course of the dilutions which 
precede spreading for assay on the petri dishes. This phenomenon shall be described and 
discussed in detail in a communication which shall appear soon in this periodical. We thus 
consider that the experiments of Price do not bring any proof of liberation of bacteriophages 
in the absence of bacterial lysis. 



325 



INVESTIGATIONS ON A LYSOGENIC BACILLUS MEGATERIUM 

The results presented below correspond to a study of 104 groups of bacteria: 

Number of drops 104 

Initial number of bacteria 295 

Final number 947 

Number of drops where lysis was observed 3 

Number of lysed bacteria 4_|_24-8 

Total of bacteria lysed I4 

The number of bacteria used was relatively restricted, but one can nevertheless 
infer that, in certain experiments, the proportion of bacteria producing phage was 
less than 5%, while in others, this proportion represented a quarter, or even a 
third, of all the bacteria used. 

Discussion. The Mutation Hypothesis. The fact that only a certain proportion 
of bacteria produced bacteriophages poses the problem of the nature of the factors 
which induced a lysogenic bacterium to produce bacteriophages. If a sensitive 
mutilat is mixed with bacteriophages, one observes the lysis of some of the cells 
and the growth of resistant lysogenic colonies. The statistical analysis of the 
phenomenon, such as has been carried out by M. Dulbriick and S. E. Luria 
(1942), for E. coli, has not been carried out, and it is impossible to state that the 
resistant bacteria represent spontaneous mutants. This is possible, or even 
probable. One could envisage the hypothesis that this mutation from sensitiv- 
ity to resistance is reversible. In the case of a lysogenic bacterium the mutation 
from resistance to sensitivity would create conditions which permit "inactivation" 
of the potential bacteriophage and of the lysis of the bacterium. One, thus, 
conceives the idea that the mutation could intervene in determining the 
production of bacteriophages by certain lysogenic bacteria. 

The mutation rate is generally independent of the conditions of the medium, 
and we have seen that the percentage of bacteria which lyse varies from .5 to 
30%. But this does not necessarily exclude the mutation hypothesis. It is, in 
fact, conceivable that this variability is the result of the simultaneous coming 
into play of a high mutation rate and of selection. We would invoke the clonal 
character of the lysis in support of this hypothesis. Let us recall that we never 
observed the lysis of a single member of a diplo-bacillus, and that we have 
sometimes seen the lysis of all of the bacteria in the same filament. This could be 
interpreted in the following manner: Only the descendants of certain mutants 
lysed, while there occur from 1 to 4 divisions between mutation and lysis. This 
hypothesis is not incompatible with the hypothesis of induction. One could see, 
in effect, that the mutation affects only a genotypic factor, and that the latter 
does not express itself by the production of bacteriophages except under specific 
conditions of the medium. It is, however, the study of the kinetics of the 
production of bacteriophages, which shall now be presented, which allows one to 
exclude the hypothesis of the intervention of a mutation. 

The Induction Hypothesis. The behavior the B. megaterium in microdrops has 
given us the impression that the production of bacteriophages must depend, in 
part, on the previous history of the bacteria, and the hypothesis was envisaged 



326 



ANDRE LWOFF AND ANTOINETTE GUT MANN 

that the inducing factor intervenes in the course of the development of the 
mother culture. The drop technique, in spite of long and laborious experiments, 
did not bring the solution to this problem, which was then taken up with ordinary- 
cultures. Since that time, experiments (A. Lwoff, L. Siminovitch, and N. 
Kjeldgaard) have shown that it is possible to achieve conditions which induce the 
lysis of 20 to 30% of a bacterial culture and that this lysis is accompanied by the 
liberation of bacteriophages. These experiments shall be described in a separate 
communication. We shall content ourselves here with a discussion of the 
conclusions to which these experiments led, that is to say, that the production of 
bacteriophages is induced by external factors. 

The Hypothesis of Burnet and McKie. F. Burnet and M. McKie (1929b) 
realized that all bacteria of the lysogenic strain contain in their hereditary 
constitution a unit which is potentially capable of liberating bacteriophages; 
that these bacteriophages can be liberated during the process of normal growth; 
that activation of the hereditary structure must take place spontaneously for 
there to be any liberation of bacteriophages; that the data do not allow one to 
infer whether or not there is an intracellular activation; that the conditions are 
probably analogous to those which intervene in monomolecular reactions or in 
radioactive disintegrations. Undoubtedly, some of these concepts did not stand 
up to the test of time. But the problem was remarkably well stated. And the 
hypothesis of "activation" should have merited better than a complete indif- 
ference. 

Everything seems to indicate that the bacteriophage possesses a remarkable 
attractive power. It inhibits, in effect, bacterial growth, detours to its own end 
the metabolites necessary for enzymatic adaptation (J. Monod and E. Woll- 
man, 1947) and growth of a different bacteriophage (M. Delbriick and 
S. E. Luria, 1942; A. D. Hershey and R. Rotman, 1948). H. J. MuUer (1947), 
while discussing the problem of the reproduction of the gene, concluded that only 
the phenomena of long range forces could account for the duplication. Perhaps 
the attractive power of bacteriophages is of this same type. 

Let us suppose that in a lysogenic bacterium the probacteriophage, under 
whatever form it might exist, is in competition with a certain number of other 
particles. In potentially lysogenic bacteria that equilibrium is stable. But if, 
for one reason or another, the equilibrium is perturbed possibly because the 
attractive power of the bacterial particles is diminished or because the attractive 
power of the probacteriophage is increased, the latter will develop and give rise 
to bacterial virus particles. More concretely and simply, one might envisage 
that the "key enzyme" necessary for bacteriophage synthesis, which M. Delbriick 
and S. E. Luria (1942) have postulated in order to explain the phenomenon of 
interference, becomes available; perhaps, because the normal substrate of this 
enzyme ceases to be synthesized. Let us note in passing that this defect of sub- 
strate synthesis could as well be the consequence of a mutation as a change of 
the conditions of the medium. In any case, the association or combination of this 
key enzyme with the probacteriophage would then set off the process of mul- 
tiplication. 



327 



INVESTIGATIONS ON A LYSOGENIC BACILLUS MEGATERIUM 

We can now discuss the "activation" theory of Burnet and McKie. The 
activation could be the consequence of a mutation. But since, as in the case of B. 
megaterium, this activation is induced by conditions of the medium, it is simplest 
to envisage, at least in the case of the bacterium which we have studied, that the 
activation is the result of the suppression of a competition between the pro- 
bacteriophage and a specific substance or particle. This is obviously but a 
working hypothesis. 

According to unpublished observations of A. Lwoff, Louis Siminovitch, and 
N. Kjeldgaard, the length of the latent period of the bacteriophage of B. meg- 
aterium in the sensitive strain under the conditions of these experiments is 45 
minutes. Thus, in the experiments of these authors, 45 minutes elapsed between 
the moment of the inflection of the bacterial growth curves and the start of the 
production of the bacteriophage, which lasted for 20 to 30 minutes. "It seems," 
they write, "as if a factor, or an ensemble of factors, among which we know that 
the aeration of the medium plays an important role, determines the condition 
which sets off the production of the bacteriophages. This sudden drop in the 
rate of growth is evidently related to the development of the bacteriophage, 
but one does not know whether it is the effect or the cause or both cause and 
effect." A. Lwoff, L. Siminovitch, and N. Kjeldgaard envisage that the factors of 
induction act by modifying the relative speed of multiplication of the bacterium 
and of the probacteriophage. The equilibrium is modified in an irreversible man- 
ner once the number of probacteriophages surpasses a critical value. In order 
to explain its clonal characteristics, it suffices to suppose that lysis does not take 
place until several divisions have occurred. The problem of the nature of the 
factors of induction and of the mechanism of their action shall be discussed in 
a future communication which shall appear in these Annates . 

VIII. Definition of Lysogenic Bacteria 

Since the liberation of bacteriophages by the lysogenic strain 899 of B. meg- 
aterium occurs by bacterial lysis, there exists an incompatibility between the pro- 
duction of bacteriophages on the one hand and the survival and, a fortiori, 
perpetuation of the bacterial individual on the other hand. In a "lysogenic" 
strain, there are bacteria which multiply and other bacteria which produce and 
liberate bacteriophages. 

A lysogenic culture producing bacteriophages can be transferred indefinitely, 
but only under the condition that it includes bacteria which do not produce 
bacteriophages. On the contrary, a bacterium producing bacteriophages is a 
bacterium condemned to death. It is thus appropriate to distinguish between a 
lysogenic culture and a lysogenic bacterium ; in case of the latter the production of 
bacteriophages is but a potential faculty. For that reason we have proposed the 
following definition: "A lysogenic bacterium is a bacterium which perpetuates 
the capacity to form bacteriophages without intervention of exogenous bacterio- 
phages." 

In the case of lysogenic bacteria, the bacteriophage perpetuates itself in the 
form of the probacteriophages and not in the form of bacteriophage virus par- 



328 



ANDRE LWOFF AND ANTOINETTE GUTMANN 

tides. One can thus compare a lysogenic bacterium to a healthy carrier of germs. 
What constitutes the particularity of lysogenic bacterium is that they per- 
petuate a potentially lethal character. The expression of this character, which 
can be induced by exogenous factors, involves the death of the bacterium. 



• bacteriophage 
o probacteriophage 




C^'y-induction(J^^^nkction.^r^^ 

*T I \ 




• • . :• ••• 



CZIJ 

J 



c^ 




Diagram showing the evolution of bacteriophage and probacteriophage 
{prophage) in sensitive and lysogenic megaterium bacilli. 

Both in the case of the sensitive bacterium, as well as in the case of a resistant 
bacterium, the penetration of the bacteriophage is followed by a disappearance 
of the infectious character. In both cases, the infectious corpuscle is not re- 
generated until a short time prior to the bacterial lysis. In both cases, finally, 
the multiplication of the bacteriophage involves the lysis of the bacterium. 
But in the case of a sensitive bacterium the penetration of the bacteriophage and 
its transformation into a non-infectious form is followed immediately by an 



329 



INVESTIGATIONS ON A LYSOGENIC BACILLUS MEGATERIUM 

important process of multiplication involving the death of the bacterium. In 
contrast, after the penetration of the bacteriophage into a resistant bacterium 
destined to become lysogenic the probacteriophage behaves as a nonpathogenic, 
normal cellular unit, which does not assume a pathological development except 
as a consequence of an "activation." This development in a bacterium of a 
particle which ultimately becomes "malignant" recalls in some respect the 
multiplication of neoplastic cells in a cancerous organism. 

One could be tempted to write, as perhaps we ourselves have done, that 
potentially lysogenic bacteria live "in equilibrium" with their bacteriophage. 
But, in fact, they live in equilibrium with a probacteriophage. There exists an 
incompatibility between the structure "bacteriophage-virus" and the survival 
of the bacterium. 

It is true that lysogenic bacteria are resistant to the bacteriophages which 
their sisters produce. But no more than a sensitive bacterium does a lysogenic 
bacterium resist the multiplication of bacteriophages. It is thus the multiplica- 
tion of bacteriophages which constitute the pathologic fact. 

IX. Resume and Conclusions 

1. Bacillus megaterium cells washed and isolated in microdrops were utilized 
in an attempt to resolve some of the problems posed by lysogenic bacteria. 

2. B. megaterium can multiply without liberating bacteriophage: production 
of bacteriophages thus does not obligatorily accompany the growth and the 
division of the bacteria. 

3. The capacity to produce bacteriophages was maintained in the course of 
19 divisions in the absence of all free bacteriophages: the faculty to produce 
bacteriophages is thus perpetuated intrabacterially. 

4. Rapid lysis of lysogenic bacteria was observed under the microscope. This 
lysis is followed, and never preceded, by liberation of bacteriophages. B. 
megaterium thus liberates bacteriophages by lysis. 

5. The proportion of bacteria which liberate bacteriophages varies within 
considerable proportions: from less than 5% to 30%. All indications are that the 
production of bacteriophages is induced by external factors. 

6. Lysogenic bacteria are defined as bacteria in which the capacity to produce 
bacteriophages is perpetuated without intervention of exogenous bacteriophages. 

7. The theory of the activation of an intracellular anlage of the bacteriophage 
(F. Burnet and M. McKie) has been discussed, developed, and amended. 



BIBLIOGRAPHY 

Adams (M. H.). /. Immunol., 1949, 62, 505-516. 

Bayne-Jones (S.) et Sandholzer (L. A.). /. exp. Med., 1933, 57, 279-304. 

BoRDET (J.). Ces Annales, 1925a, 39, 717; C. R. Soc. Biol. Beige, 19256, 93, 1054. 



330 



ANDRE LWOFF AND ANTOINETTE GUTMANN 

BoRDET (J.) et Renaux (E.). Ces Annales, 1928, 42, 1283. 

Burnet (F. M.) et Lush (D.). Austr. J. exp. Biol., 1936, 14, 27-38. 

Burnet (F. M.) et McKie (M.). Austr. J. exp. Biol., 1929a, 6, 21-31 et 19296, 6, 277-284. 

FoNBRUNE (P. de). Technique de micromanipulation. Masson, edit., Paris, 1949. 

DELBRtJCK (M.). Biol. Rev., 1946, 21, 30-40. 

Delbruck (M.) et Luria (S. E.). Arch. Biochem., 1942, /, 111-141. 

Den Dooren de Jong (L. E.). Zentralbl. Bakt., 1931a, 120, 1-23; Ibid., 19316, 122, 

277-286; Ibid., 1934, 131, 401-410. 
DuGGAR (B. M.) et Armstrong (J. K.). Ann. Missouri Bot. Garden, 1923, 10, 191. 
Foster (R.). J. Bad., 1948, 56, 795-809. 

Gorini (L.) et Fromagenot (C). C. R. Acad. Sci., 1949, 229, 559-561. 
Gratia (A.). Ces Annales, 1936a, 56, 307; C. R. Soc. Biol. Beige, 19366, 122, 812; Ibid., 

1936c, 123, 506; ces Annales, 1936d, 57, 652. 
Gratia (A.) et Welch (M.) C. R. Soc. Biol. Beige, 1939; 132, 330. 
Hershey (A. D.). Genetics, 1946a, 31, 620-640; Cold Spring Harbor Symp. Quant. Biol., 

19466, /;, 67-77. 
Hershey (A. D.) et Bronfenbrenner (J.). In "Viral and rickettsial infections of man." 

Lippincott, edit., 1948. 
Hershey (A. D.) et Rotman (R.). Proceed. Natl. Acad. Sci., 1948, 34, 89-96; Genetics, 

1949, 34, 44-71. 
Latarjet (R.). J. gen. Physiol, 1948, 31, 529-546. 
Luria (S. E.). Proceed. Nat. Acad. Sci., 1948, 33, 253-264. 
Luria (S. E.) et Delbruck (M.). Genetics, 1943, 28, 491-511. 
Luria (S. E.) et Dulbecco (R.). Genetics, 1949, 34, 92-125. 
Luria (S. E.) et Latarjet (R.). J. Bad., 1947, 53, 149-163. 

Luria (S. E.) et Palmer (J. L.). Carnegie Inst. Washington Yearbook, 1945-1946, 154. 
LwoFF (A.). In "Unites biologiques douses de continuity g^n^tique," C.N.R.S., 6dit., 

Paris, 1949, 7-24. 
LwoFF (A.) et Gutmann (A.). C. R. Acad. Sci., 1949a, 229, 605-607; Ibid., 19496, 229, 

679-682: Ibid., 1949c, 229, 789-791; Ibid., 1950, 230, 154-156. 
LwoFF (A.), Siminovitch (L.) et Kjelgaard (N.). C. R. Acad. Sci., 1950, 230, 1219; 

Ces .Aimales (sous presse). 
McLaughlan (T. a.), Clark (E. M.) et Boswell (F. W.). Nature, 1947, 160, 755-756. 
MoNOD (J.) et AuDUREAU (A.). Ces Annales, 1946, 72, 868. 
MoNOD (J.) et WoLLMAN (EUe). Ces Annales, 1947, 73, 937. 

MuLLER (H. J.). Am. Naturalist, 1922, 56, 32; Proceed. Roy. Soc. B., 1947, 134, 1-37. 
Northrop (J. H.). /. gen. Physiol, 1939, 23, 59-79. 
PiRiE (A.). Brit J. exp. Path., 1940, 21, 125-132. 
Price (W. H.). J. gen. Physio., 1949, 32, 481-488. 
Rhoades (M. M.). In "Unites biologiques douses de continuity g^n^tique," C.N.R.S., 

6dit., Paris, 1949, 37-44. 
WoLLMAN (E.) et (E.). C. R. Soc. Biol, 1936a, 121, 126; Ibid., 19366, 122, 190; Ibid., 

1936c, 122, 871; Ces Annales, 1936(i, 56, 137; Ibid., 1936e, 56, 316: C. R. Soc. Biol, 

1937, 124, 931; Ces Annales, 1938, 60, 13. 



331 



MICROBIOLOGY 

INDUCTION OF BACTERIOPHAGE LYSIS OF AN ENTIRE 
POPULATION OF LYSOGENIC BACTERIA' 

by 

Andre Lwoff, Louis Siminovitch, and Niels Kjeldgaard 

Presented by Robert Courrier 

In a culture of lysogenic Bacillus megaterium, some bacteria multiply 
without liberating bacteriophages and perpetuate the lysogenic strain, 
while others produce bacteriophages which are liberated by lysis. ^ Under 
normal conditions of exponential growth, only a very small percentage 
of bacteria produce bacteriophages.^'^ 

We have succeeded in inducing bacteriophage lysis of an entire popula- 
tion of lysogenic bacteria by means of irradiation by ultraviolet light. 

B. megaterium is grown in a yeast extract medium.- During the expo- 
nential growth phase, when the number of bacteria has attained 34 X 10® 
per ml., a 2 mm layer of the culture is irradiated with ultraviolet light from 
a high pressure mercury vapor lamp, which delivers to the surface of the 
liquid an energy of 2000 ergs per mm- per minute of radiation of wave 
length 2537 A. The culture is then shaken at 37° and its optical density 
(O.D.) increases for about 80 minutes. The normal growth rate, that is 
to say the number of doublings per hour, is 3. Immediately after irradiation 
the growth rate is in the neighborhood of L5. At about the 80th minute, 
when the O.D. has increased by a factor of 3 to 4, bacterial lysis takes 
place: the culture clears in 40 to 80 minutes. The fraction of bacteria 
surviving is usually less than 10~*. Lysis is accompanied by the liberation 
of about 70 to L50 bacteriophages per bacterium. Analogous results have 
been obtained with irradiation for 20, 30, 90, or 120 seconds. 

A culture growing in synthetic medium and similarly irradiated from 1 to 
60 seconds does not lyse. On the other hand, if a culture growing in yeast 
medium is centrifuged, resuspended in synthetic medium, and then ir- 
radiated for 5, 10, 20, or 30 seconds, bacteriophage lysis does take place. 

Translated from the French and reprinted by permission of the 

authors and the Imprim^rie Gauthier-Villars from the Comptes 

Rendus des Seances de l'Academie des Sciences, 231, 

190-191 (1950). 

332 



Microbiology 

Finally if yeast extract is added to a culture growing in synthetic medium 
and the culture irradiated immediately for 10, 20, 30, or 60 seconds, no 
lysis takes place. But irradiation carried out 20 to 40 minutes after addition 
of the yeast extract (by which time the O.D. has increased by about 50%) 
induces bacteriophage lysis. UV irradiation thus induces production of 
bacteriophages only with bacteria which have grown for 20 to 40 minutes 
in a complex organic medium such as yeast extract. 

Irradiation for 30 to 60 seconds of a non-lysogenic culture of B. mega- 
terium growing in yeast extract does not appear to affect bacterial growth. 
The lysis of the lysogenic strain is thus probably not a direct effect of the 
UV irradiation. 

It has been demonstrated previously that all bacteria of a lysogenic 
strain are capable of perpetuating the lysogenic character. ^ The experi- 
ments described here, which have been complemented by studies of bacteria 
irradiated and then isolated in microdrops, now demonstrate that under 
certain conditions, all the bacteria of a lysogenic population are capable 
of undergoing lysis with liberation of bacteriophages. 



'Supported by a grant of the National Institutes of Health of the United States of 
America. 

2A. Lwoff and A. Gutmann, Ann. Inst. Pasteur, 78, 711-739, (1950). 

*A. Lwoff, A. Siminovitch, and N. Kjeldgaard, C'omptes rendus, 230, 1219-1221, (1950); 
Ann. Inst. Pasteur, 79, 815 (1950). 



333 



GENETICS 

LYSOGENY AND GENETIC RECOMBINATION 
IN ESCHERICHIA CO LI K12 

by 

Elie L. Wollman and Frangois Jacob 

Presented by Jacques Trefouel 

Crosses between E. coli K12 Hfr and F^, both Ij'sogenic, show that the X pro- 
phage and Gal 4 are situated on the segment TL LaCj. Spontaneous induction 
of the prophage suppresses those of the zygotes which received a X prophage 
whenever the Hfr parent is lysogenic and the F" parent non-lysogenic. This 
modifies the distribution among the recombinants of characters Unked to the 
X prophage. 

It had been supposed until now that the presence of the X prophage 
does not alter the results of genetic analysis of E. coli K12. The phe- 
nomenon of spontaneous induction observed in the course of crosses 
between lysogenic Hfr (ly+) bacteria and non-lysogenic F~ (ly~) 
bacteria established the fact that in the course of certain crosses an im- 
portant fraction of the zygotes can be eliminated.' One might ask whether 
this elimination does not affect preferentially some segments of genetic 
material, thus exercising a counterselection with respect to certain classes 
of recombinants. In particular, it might be possible to explain in this way 
the anomalies found in crosses of the type F+ ly+ X F~ ly~.^ 

In crosses of Hfr T+ L+ S^ X F" T" L" S^ in which the bacteria are 
mixed in the proportion of 1 Hfr bacterium for 20 F~ bacteria the fre- 
quency of recombination (the number of T+ L+ S'' recombinants divided 
by the initial number of Hfr) varies considerably according to the actual 
cross carried out. This frequency, which is of the order of 5 to 10% for 
crosses Hfr ly+ X F~ ly~, surpasses 50% in crosses of Hfr ly+ X F~ 
ly+ and of Hfr ly- X F" ly". 

Translated from the French and reprinted by permission of the 

authors and the Imprim^rie Gauthier-Villars from the Comptes 

Rendus des Seances de l'Academie des Sciences, 239, 

455-456 (1954). 

334 



Genetics 

The distribution of non-selected markers among the recombinants is also 
affected by the nature of the crosses. In a cross of Hfr ly~ Lac^ Gal+ X 
F~ ly~ Lac~ Gal~ 55% of the recombinants are Lac"^, 45% are Gal"*". 
The simultaneous passage of lac+ and Gal+ occurs in 40% of the cases, 
which indicates linkage between Lacj and GaU. In a cross of Hfr ly"*" 
Lac+ Gal| X F~ ly~ Lac^ Gal~, in contrast, 25% of the recombinants 
are Lac,+ but only 3% are Gal"*", of which about half are also Lac"*". Among 
these recombinants, only 0.2% are lysogenic. 

It is known that there exists a direct linkage between lysogeny and the 
character GaU.^'- The extreme rarity of lysogenic recombinants, as well 
as the considerable diminution of the proportion of Gal+ recombinants, 
is the consequence of the spontaneous induction of the prophage in a 
large fraction of the zygotes. The relatively smaller diminution of the 
proportion of Lac"*" recombinants similarly reflects this phenomenon, 
but shows at the same time that the character Lacj is much less linked to 
the characters Gal 4 and ly than the latter are linked to each other. LaCi 
is apparently located between TL and Gal 4. The localization of the 
prophage in this region of the bacterial chromosome is demonstrated by 
crosses of the type Hfr ly+ Lac+ TP X F" ly+ Lac^ TV, in which the 
two lysogenic strains perpetuate different X prophages. The prophage of 
the Hfr parent is transmitted to 30% of the recombinants while the Lac 
and the TP characters are each transmitted in 50 and 75% of the cases. 

It should be noted that the establishment of the prophage of the Hfr 
parent in the recombinants is one third as frequent as the development 
of this prophage in the zygote in the course of crosses Hfr ly"*" X F~ ly~. 

It is known that in a cross of Hfr X F~ the high frequency of recombina- 
tion observed concerns only the segment TL Lacj. We have verified this 
fact in all of our crosses. The relatively high frequency of recombination of 
the character X lysogeny thus localizes the X prophage in this segment. 

The elimination of a chromosomal segment by spontaneous induction 
of the X prophage could also occur, though to an extent difficult to evaluate, 
in crosses of the type ly+ X ly+ or F+ ly" X F" ly+. It offers a 
model which might conceivably explain the segmental eliminations des- 
cribed by J. Lederberg in heterozygote diploids^ and by W. Hayes* in 
recombinants of Hfr strains. 



IF. Jacob and E. L. Wollman, Comptes rendus, 239, 317, (1954). 
2E. L. Wollman, Ann. Inst. Pasteur, 84, 281, (1953); R. K. Appleyard, Cold Spring 
Harbor Symp., 18, 95, (1953). 

3E. M. and J. Lederberg, Genetics, 38, 51, (1953). 
4W. Hayes, Cold Spring Harbor Symp., 18, 75, (1953). 
5J. Lederberg, Proc. Natl. Acad. Sci., 35, 178, (1949). 



335 



GENETICS 

SPONTANEOUS INDUCTION OF THE DEVELOPMENT 

OF BACTERIOPHAGE X IN GENETIC RECOMBINATION OF 

ESCHERICHIA CO LI K12 

by 

Frangois Jacob and Elie WoUman 

Transmitted by Jacques Trefouel 

In the course of crosses between lysogenic Hfr and non-lysogenic F" 
bacteria of E. coli K12, the prophage passes from the lysogenic to the non- 
lysogenic parent becomes induced, and develops in the latter. This process 
leads to the destruction of the zygote formed. 

In crosses between lysogenic (ly"*") and non-lysogenic (ly~) Escherichia 
coli K12 we have not been able to demonstrate the transfer of the lysogeny 
of the F+ parent to a recombinant possessing essentially F~ genome.^ 
Nevertheless, the lysogenic character segregated and appeared to be linked 
to another genetic factor. '^> ^ Since the X prophage is inducible, it appeared 
possible that recombination between F+ly+ and F~ly" involves the 
development of the phage and the destruction of the zygotes formed' 
The discovery of strains giving a high frequency of recombination (Hfr)' 
makes it possible to verify this hypothesis. 

Hfr ly+ S^ bacteria (lysogenic and sensitive to streptomycin) and F~ 
ly"/X S'' bacteria (non-lysogenic and resistant to X and to streptomycin) 
are mixed in broth in the proportion of 1 Hfr to 20 F~ . This mixture is 
shaken at 37° and at various times samples are removed and spread on strep- 
tomycin containing agar seeded with indicator bacteria resistant to strep- 
tomycin but sensitive to phage X. A control culture of Hfr is treated in the same 
manner. 

One observes that the number of infective centers in the mixture in- 
creases linearly as a function of time until about the 40th minute, when a 
plateau is reached which corresponds to approximately half of the initial 



IE. L. Wollman, Ann. Inst. Pasteur, 84, 281, (1953). 
2E. M. and J. Lederberg, Genetics, 38, 51, (1953). 

3L. L. Cavalli and H. Heslot, Nature, 164, 1057, (1949); W. Hayes, Cold Spring Harb. 
Symp. 18, 75, (1953); W. Haj'es has sent us his strains of Hfr. 

Translated from the French and reprinted by permission of the 

authors and the Imprim^rie Gauthier-Villars from the Comptes 

Rendus des Seances de l'Academie des Sciences, 239, 

317-319 (1954). 

336 



Frangois Jacob and Elie Wollman 

number of Hfr. This number is 100 times greater than the number of 
infective centers found in the control. After the 60th minute, the number 
of infective centers suddenly increases and at about the 150th minute 
reaches a value 50 to 100 times greater than that of the first plateau. 

The difference between the mixture and the control Hfr is further ac- 
centuated if anti-phage serum and streptomycin are added to the two cul- 
tures at the beginning of the experiment. For, the number of infective 
centers in the control falls rapidly to zero while the number of infective 
centers in the mixture is only reduced by one-half. 

The development of phage X has thus been induced by putting Hfr ly+ 
in contact with F~ ly~. This development can only take place in F~ 
bacteria (or in the zygotes), since the Hfr bacteria were sterilized by 
streptomycin. If the same cross is repeated with an F~ S^ strain, addition 
of streptomycin during the latent period arrests the development of the 
phage. 

An even more direct demonstration of the passage of the prophage from 
Hfr to F" bacteria (or to the zygote) and of its development in the latter 
is furnished by the following experiment. It is known that certain bacterial 
strains confer a phenotypic modification to the bacteriophages which they 
produce.'* Thus bacteriophage X(K12) produced by E. coli K12(X), 
whose efficiency of plating is identical on E. coli K12 S and on E. coli C, 
has a greatly reduced efficiency of plating (10~' to 10~^) on K12 S after 
passage on E. coli C. If development of the prophage actually takes place 
in the non-lysogenic F~ bacteria, then in a cross of K12 Hfr(X) S^ with 
C/X F~ S'' the bacteriophages produced should be X(C) and not X(K12). 
Experiments show this to be the case. 

When F~ bacteria perpetuate a prophage, normal or defective, the 
immunity which the prophage confers on them* protects them equally 
against the development of a prophage coming from the Hfr parent. 

If one crosses Hfr bacteria superinfected with virulent mutants of the 
type Xv,* one observes little or no passage of the superinfecting phages into 
the F~ bacteria. This result indicates that the transfer of phage material 
from Hfr to F~ takes place in the prophage state and not in the vegetative 
state. It also indicates that the induction of phage development in the 
course of recombination takes place in the F~ bacterium (or in the zygote) 
and not in the Hfr bacterium. 

It is evident that the effect described here must introduce important 
distortions in the genetic analysis of E. coli K12: in effect, only those zygotes 
give rise to recombinants in which no multiplication of the phage has 
taken place. The proportion of these survivors will vary considerably 
depending on the nature of the cross. Thus in the cross of Hfr ly"^ Bj S^ 
by W678 T~ L~ B~ ly~/XS'^ the number of recombinants of the type 



*G. Bertani and J. J. Weigle, J. Bad., 65, 113, (1953). 

T. Jacob and E. L. Wollman, Cold Spring Harb. Symp., 18, 101, (1953). 



337 



Genetics 

T+ L+ S"^ does not exceed 5% of the initial number of Hfr bacteria, 
although more than 50% of these bacteria give rise to infective centers. 
Among 200 recombinants examined not one was lysogenic. In a compar- 
able cross in which the F~ strain perpetuates a defective prophage, the 
number of T+ L"*" S"" recombinants reaches 50% of the initial number of 
Hfr, although the number of infective centers does not surpass 5%. 
Among the recombinants about 30% are normal lysogenics. 

The phenomenon of spontaneous induction of the prophage is also found 
in crosses of F+ ly+ by F~ ly~ and thus explains the observed absence 
of the transmission of lysogeny in these crosses. It could also be the origin 
of other anomalies found in the study of the genetics of E. coli K12. 



338 



Transduction of Lysogeny in Escherichia co/i 

Francois Jacob 

Service de Physiologie microhienne, Institut Pasteur, Paris 
Received, March 2, 1955 

SUMMARY 

Transduction provides a new tool for a genetic analysis of lysogeny. Re- 
cently, transduction has been shown by Lennox to occur, through phage PI, in 
various strains of Escherichia coli. Another strain of temperate phage has been 
isolated, which is also able to transfer genetic characters from a donor to an 
acceptor strain of E. coli K12. Linked characters can be transduced simul- 
taneously. Lysogeny or nonlysogeny, with respect to each of three different 
prophages is transduced together with a galactose marker, to which these 
three prophages are linked as shown by bacterial recombination evidence. 
When two geneticallj^ different and complementary prophages — one in the 
donor and one in the acceptor cells — are used in transduction experiments, 
recombination of prophages has been shown to occur. 

INTRODUCTION 

A new mechanism allowing the transfer of genetic characters from 
one bacterial strain to another was found in Salmonella by Zinder and 
Lederberg (1952). This mechanism, for which the term "transduction" 
was coined, appears to involve phage particles as vectors of the genetic 
material of the bacteria (Zinder, 1953; Stocker et al., 1953). Transduction 
can be observed when phage-sensitive bacteria, acting as acceptor cells, 
are infected with phage particles grown on donor bacteria which differ 
from the acceptor by one or several genetic characters. Some of the in- 
fected bacteria that survive phage infection acquire and transmit to 
their progeny one or several characters of the donor strain. The trans- 
ducing ability seems to be restricted to some strains of temperate phages. 

It has recently been shown by Lennox (1955) that phage PI is able 
to transduce various markers from one strain of E. coli to another and 
that linked characters can be transduced together. This finding makes 
it possible to compare linkage data provided by transduction and by 
bacterial recombination which was demonstrated by Tatum and Leder- 
berg (1947) to occur in strain K12 of E. coli. 



Reprinted by permission of the author and Academic Press, Inc. 
from Virology, 1 (2), 207-220 (1955). 



339 



208 FRANgOIS JACOB 

This paper is mainly concerned with preliminary results on trans- 
duction of lysogeny in E. coli K12 by the use of a strain of temperate 
phage isolated in our laboratory. Previously, information on the genetics 
of X lysogeny was gained through bacterial recombination experiments 
with E. coli K12 (Lederberg and Lederberg, 1953; Wollman, 1953; 
Appleyard, 1954a; Wollman and Jacob, 1954). The results indicated 
that the X-lysogenic character segregates in crosses between lysogenic 
and nonlysogenic bacteria and, moreover, that this character is linked 
to a marker that plays a role in galactose utilization. Nevertheless, 
whereas the presence of a nuclear unit controlling lysogeny was thus 
demonstrated, it had not yet been proved fully that this bacterial unit 
was identical to the prophage, i.e., to the genetic material of the phage 
in the lysogenic condition (Lwoff, 1953). Transduction of lysogeny offers 
a new way of analyzing the relations between prophage and bacterium, 
as well as the size and orientation of the prophage. 

MATERIAL AND METHODS 

Transducing phage. Various kinds of temperate phages have been 
tested for their transducing ability in E. coli K12. Of twenty-three 
phages released by difTerent lysogenic strains of E. coli isolated from 
patients, only one was active: phage 363. It forms tiny turbid plaques 
on E. coli K12 (Fig. lA). Its latent period in broth is about 50 minutes, 
and its burst size is 200 to 300. Its frequency of lysogenization is higher 
at 20° than at 37°. Phage 363 appears to belong to the same group as 
phage PI, previously described by Bertani and Nice (1954) and used 
by Lennox (1955) in his transduction experiments. Bacteria lysogenic 
for 363 are immune against PI, and vice versa. Phage 363 is inactivated, 
although at a lower rate than PI, by an anti-Pl serum. ^ 

Transduced prophage. The properties of phage X and its relationship 
with the bacterial host have been described (Lederberg and Lederberg, 
1953; Weigle and Delbriick, 1951). Two other temperate and ultraviolet- 
inducible phages, released by lysogenic E. coli isolated from patients, 
have also been used in these experiments: phages 82 and 434 (Fig. IB, D). 
The properties of these two phages will be described elsewhere. Phages 
82, X, and 434 can easily be distinguished by their host range, as shown 
in Table 1. 

Bacterial strains. The bacterial strain used as a donor was a proto- 
troph, K12, sensitive to 363, 82, X, and 434. 

' A sample of anti PI serum (anti H~) was kindly supplied by Dr. J. Beumer. 



340 



TRANSDUCTION OF LYSOGENY IN ESCHERICHIA COLI 209 




# 



Fig. 1. Plaques formed on E. coli K12 lj\- (A) piuige '663, (B) phage 82, (C) phage 
Xms, and (D) phage 434. 

The acceptor strain was a derivative of W677 (Lederberg, 1947) pre- 
pared by Dr. E. L. WoUman. This strain, called P678, is unable to syn- 
thesize threonine (T"), leucine (L~), or vitamin Bi (BY) or to ferment 
lactose (Laci), xylose (Xyl~), mannitol {Man~), maltose (Mal~), or 
galactose (Gait, Galh). It is not lysogenic for X (X)~ and is resistant to 
streptomycin (S"). The Gah marker of this strain is closely linked to the 
X locus. Its relationship to other galactose loci in the same region is 
unknown. 

Substrains of K12 and P678, lysogenic for one or more of the four 
prophages (363, 82, X, and 434), have been isolated. In the description 
of the following experiments, P678(X)+ will refer to a substrain of P678 
lysogenic for X. Variants resistant to one or more of the three phages 
have also been used: P678/X will refer to a variant of P678 resistant to X. 

Preliminary crosses between GalhS^Hfr, nonlysogenic, and GalhS''F~, 
lysogenic for one of the prophages, indicate that both prophages 82 and 
434, as well as X, are located in the Galh region. When S'' and Galb are 
used as selective markers, the proportion of nonlysogenic recombinants 
is about 90% with phage 82, 80% with phage X, and 70% with phage 
434. Prophage 363 is located in a different region. 



341 



210 FRANgOIS JACOB 

TABLE 1 
Host Range of Phages 82, X, and 434 







Phages 




Bacterial strain 










82 


X 


434 


K12(82)+ 





+ 


+ 


K12(X) + 


+ 





+ 


K12(434)+ 


+ 


+ 





K12/X 


+ 





+ 


K 12/82, 434 





+ 






Media. For infection experiments, broth supplemented with 10~^ M 
CaCU was used. 

The selective medium was the following: KH2PO4 — 13.6 g; (NH4)2S04 
— 2 g; Ca (NOOz— 0.001 g; MgS04-7 H2O— 0.02 g; Difco agar— 20 g; 
H2O — 1000 ml; KOH added to pH 7.0. Sugars were added separately at 
a concentration of 1 %. Amino acids were added at a concentration of 
100 Mg/ml; vitamin Bi at 5 Mg/ml. 

Transduction experiments. Phage 363 was grown on the prototroph 
strain K12, either lysogenic or not. The preparations were sterilized 
with chloroform. Their titer usually reached between 2 and 8 X 10^ 
particles per milliliter. 

For transduction experiments, about 5 X 10^ particles of phage 363 
were added to 4 ml of a broth culture containing about 2X10^ bacteria 
per milliliter of the acceptor strain, P678 aS''. This mixture was shaken 
for 2 hours at room temperature and then centrifuged and washed. The 
bacterial pellet was resuspended in buffer, and aliquots were plated on 
selective minimal medium supplemented with streptomycin to avoid 
contamination. In each experiment, the phage preparation was tested 
for sterility as well as for plaque count, and a noninfected culture was 
assayed as a control for spontaneous mutants. 

EXPERIMENTAL RESULTS 

Evidence for transfer of bacterial genetic material through phage 363. 
Phage 363 can transfer various kinds of characters from donor to ac- 
ceptor cells. As previously found by Lennox (1955), unlinked characters 
are transduced independently, whereas characters which are known to 
be hnked can be transduced simultaneously. In Table 2 are reported 
the results of an experiment in which several characters have been trans- 



342 



4 


730 


1 


1280 





23 





63 





129 



TRANSDUCTION OF LYSOGENY IN ESCHERICHIA COLI 211 

TABLE 2 
Transduction of Various Markers with Phage 363 

nor 's ft"v^i ns Colonies per 5.6 X W 

Selection for '^.tninflrt J bacteria infected with 

bacteda 1.2 X 10« phage 363 

Threonine"'" 

Leucine"*" 

Threonine"*" + leucine^ 

Lactose"*^ 

Galactoseb 

Acceptor P678 bacteria infected with phage 363 grown on prototrophic K12 
(multiplicity of infection, 0.3), and uninfected control bacteria were shaken for 
2 hours at room temperature. After centrifugation, the bacteria were resus- 
pended in buffer and plated on various selective media. 

duced. Aliquots of noninfected P678 bacteria and of bacteria infected 
with phage 363 grown on a prototrophic strain were plated on various 
selective media. It is seen in Table 2 that the nutritional characters 
threonine^ or leucine"+^ are transferred separately with a frequency of 
about 10~^ per adsorbed particle. These two markers, which are known 
to be linked from bacterial recombination experiments (Lederberg, 
1947), can be transferred simultaneously with a frequency of about 10^'' 
per phage particle. 

The two fermentative markers, Lac^ and Gait, are transmitted with 
a lower efficiency than T^ or L+. Gait, with which the experiments re- 
ported in this paper will be mainly concerned, is transferred with a fre- 
quency of about 0.7 to 1.5 X 10~^ per phage particle. Four hundred 
colonies selected for transduction of Gait were assayed for threonine, 
leucine, lactose, mannitol, and xylose. None of them had acquired any 
of these characters other than galactose utilization. 

The efficiency of transduction varied widely from one experiment to 
another and from one preparation of phage 363 to another. In various 
experiments, the frequency of lysogenization with phage 363 in the 
transduced clones was found to vary from 30 to 70%. Nevertheless, 
in a given experiment, this proportion was about the same for each class 
of transduced cells, independently of the selective marker. 

Transfer of nonlysogeny. In order to demonstrate transfer of non- 
lysogeny, bacteria lysogenic for one of the three prophages — 82, X, or 
434 — and resistant (nonabsorber) to the homologous phage were infected 



343 



212 



FRANgOIS JACOB 



TABLE 3 
Transfer of NOiXLYSOGENY 





Acceptor strain 


Selection for 


Experi- 


Threonine* 


Lactose* 


Galactose 


^ 


No. 


Number 

of 

colonies 

tested 


Number 

of 

non- 

lysogenic 


Number 

of 
colonies 
tested 


Number 

of 

non- 

lysogenic 


Number 

of 
colonies 
tested 


Number 

of 

nonlyso- 

genic 


% 


1 

2 


P678(82)+/82 
P678(82)+/82 


200 





200 

200 







200 
400 


19 
53 


9.5 
13 


1 
2 


P678{X)+A 
P678(X)+/X 


200 
200 






200 
200 






200 
400 


12 

32 


6 

8 


1 
2 


P678(434)+/434 
P678(434)+/434 


160 





160 





160 
400 


3 
2 


1.8 
0.5 



Acceptor bacteria P678 lysogenic for one of the three prophages — 82, X, or 434 — 
and resistant to the homologous phage were infected with a preparation of phage 
363 grown on nonhsogenic K12. After plating on various selective media, col- 
onies were tested for their ability to release the original phage, 82, X, or 434. 

with a preparation of 363 grown on prototrophic nonlysogenic bacteria. 
Infected bacteria were plated on various selective media, and colonies 
were tested for their ability to release the phage of the parent strain 
(82, X, or 434). The results of such experiments are reported in Table 3. 
It is seen, on the one hand, that among the T^, L+, or Lac^ colonies 
none was found to be nonlysogenic. On the other hand, a small fraction 
of the Gal^ clones was found to be nonlysogenic, the proportion of non- 
lysogenic clones depending on the nature of the prophage involved 
This result is in agreement with bacterial recombination experiments 
which show that each of these prophages exhibits a different linkage to 
the Galh locus. 

A question raised by the transduction of nonlysogeny is whether only 
a piece of the prophage can be replaced by an homologous region of 
nonlysogenic bacteria. Such a partial substitution could eventually result 
in an immune defective clone, i.e. a clone in which bacteria possess a 
fragment of the prophage which does not allow the synthesis of infective 
particles, but prevents phage multiplication after infection with homol- 
ogous particles. In order to detect such clones, transduction of non- 
lysogeny was also performed on lysogenic bacteria able to adsorb homolo- 



344 



TRANSDUCTION OF LYSOGENY IN ESCHERICHIA COLI 



213 



gous phage particles. With prophage X, no defective immune clone was 
found. With phage 82, it was found that 2 out of 21 Gal+ clones which 
did not produce infective phage were immune against phage 82 and its 
virulent mutant, 82c. After ultraviolet induction, these two strains ex- 
hibited a small degree of lysis, but no infective phage was released. After 
several generations, both strains had become sensitive to phage 82. These 
results suggest that recombination might have occurred between the 
prophage and an homologous genetic fragment issued from nonlysogenic 
bacteria. Unfortunately, because of the lack of markers on phage 82, no 
genetic analysis was undertaken. 

Transfer of lysogeny. In order to demonstrate transfer of lysogeny, 
nonlysogenic bacteria 678/X/82, 434 were infected with preparations of 
363 grown on prototrophic bacteria lysogenic for one of the three phages 
(82, X, or 434), and the bacteria were plated on various selective media. 
The results of such experiments are listed in Table 4. As in the case of 
nonlysogeny, lysogenic colonies were found only among those selected 
for galactose utilization. For each prophage, the frequency of transfer 
of lysogeny appears to be of the same order of magnitude as that found 
for the transfer of non-lysogeny. 

TABLE 4 
Transfer of Lysogeny 





Phage 363 
grown on strain 


Selection for 


Experiment 


Threonine^ 


Lactose* 


Galactose^ 


No. 


Number 

of 

colonies 

tested 


Number 

of 
lysogenic 


Number 

of 

colonies 

tested 


Number 

of 
lysogenic 


Number 

of 

colonies 

tested 


Number 
of 
lyso- 
genic 


% 


1 

2 


K12(82)+ 
K12(82)+ 


200 





200 





200 
400 


21 
49 


10.5 

12.25 


1 
2 


K12(X)+ 

K12(X)+ 


200 





200 





200 
400 


11 
24 


5.5 
6 


1 
2 


K12(434)+ 
K12(434)+ 


200 





200 





200 
400 


2 
1 


1 
0.5 



Acceptor nonlysogenic P678/X/82, 434 bacteria were infected with one of three 
preparations of phage 363 grown on prototrophic lysogenic strain, K12(82)+, 
K12(X)''', and K12(434)+, respectively. After plating of each mixture on various 
selective media, colonies were tested for their ability to release the correspond- 
ing phage, 82, X, or 434. 



345 



214 FRANgOIS JACOB 

In another experiment, a preparation of phage 363 was grown on a 
double lysogenic prototroph, K12(82)+(X)+, and used for infecting non- 
lysogenic P678/82/X. The colonies having acquired the Gait character 
were tested for their ability to release phages 82 or X, and some were 
found to release both. 

As has already been mentioned, in a given experiment, the proportion 
of transduced clones which are found to be lysogenic for 363 is inde- 
pendent of the selective marker. This is also true in transduction experi- 
ments involving the transfer of lysogeny for one or more of the three 
prophages — 82, X, and 434. The proportion of clones found to be lyso- 
genic for 363 is the same, whether or not the transduced clones have be- 
come lysogenic for 82, X, or 434. Since such transduction experiments were 
performed with a multiplicity of infection lower than one, it is clear that 
the transfer of lysogeny for 82, X, or 434 is performed through phage 363 
and is not the result of some kind of "phenotypic mixing." 

Reconstruction experiments show that no appreciable reinfections 
and lysogenizations with phage 363 occur on the selective plates. It 
appears therefore that the genetic material of several viruses can be 
included in a single phage coat. 

In transduction experiments, the transfer of lysogeny has been found 
to be about as frequent as that of nonlysogeny. This finding is not in 
agreement with the results found in bacterial recombination. In crosses 
between lysogenic Hfr(X)+ and nonlysogenic F", X prophage develops 
in most of the zygotes (Jacob and Wollman, 1954a). This development 
brings out the destruction of such zygotes which lyse and release phage 
particles. Lysogenic recombinants are therefore very rare. On the con- 
trary, little or no development of the prophage is observed in the reverse 
cross between nonlysogenic Hfr and lysogenic F~(X)+, and the frequency 
of nonlysogenic recombinants found agrees with expectations. The dis- 
crepancy between the results obtained by transduction and those ob- 
tained by bacterial recombination can be due to the fact that trans- 
duction experiments were performed at 20°, a temperature at which 
phages 82, X, and 434 do not multiply, whereas bacterial crosses were 
performed at 37°. It is therefore possible that, at 37°, a fraction of the 
transduced prophages is able to develop during transduction experi- 
ments. A temperature effect on the frequency of X-lysogeny transfer 
was observed by Appleyard (1954a). 

In order to test this possibility, the following experiment was designed 
(see Table 5). Phage X (as well as 82 or 434) forms plaques with the same 



346 



TRANSDUCTION OF LYSOGENY IN ESCHERICHIA COLI 



215 



TABLE 5 

Development of Transduced X Prophage 



Preparation of 363 
grown on K12(X)'^ 



Number of particles per 

milliliter forming a 363 plaque 

when plated on K12/X 



Number of particles per milliliter 
forming a \ plaque when plated on 



K12(363)-* 



K 12/363 



Untreated 1.8 X lO^ 

10 minutes after addition 1.7 X 10* 

of anti-X serum 
30 minutes after addition 1.9 X 10* 

of anti-X serum 
Same preparation, previ- 3.7 X 10" 

ously treated with anti-X 

serum, 30 minutes after 

addition of anti-363 

serum 



2.3 X 10^ 1.15 X 103 
1.2 X lO^" 1.3 X 102 

1.4 X 103 79 X 10' 
4.1 X 101 4.9 X 10' 



To a preparation of phage 363 grown on K12(X)"'', anti-X serum was added at 
time and the mixture was incubated at 37°. After 30 minutes, anti-363 serum 
was added. Before addition of any serum, and at various times during the ex- 
periment, samples were removed and diluted. Aliquots were plated as follows 
(after 30 minutes preadsorption at 37°) : ()) on K12/X (for plaques of 363) j {2) on 
K12(363)+, which adsorbs both X and 363 without multiplication of the latter 
(for total X plaques); (3) on K12/363, which adsorbs X but not 363 (for free X 
phage). Figures indicate the number of plaques found without treatment, after 
treatment with anti-X serum alone, or after successive treatments with both 



efficiency of plating on K12, K12(363)+, or K12/363. In the prepara- 
tions of 363 grown on lysogenic K12(X)+, there are always some free X 
particles which have been released during the growth of the bacteria. 
Such preparations give the same number of X plaques on K12, K12(363)"'", 
and K12/363. When most of the free X particles are neutralized by anti-X 
serum, a higher number of X plaques is found by plating on K12(363)+, 
which adsorbs 363 without allowing its multiplication, than on K 12/363 
phage, which does not adsorb phage 363. If, now, the preparation pre- 
viously treated with anti-X serum is exposed to anti-363 serum, the 
number of X plaques on K12(363)''' is reduced at a rate similar to the one 
found for inactivation of total plaques of 363. After about 30 minutes 
of exposure to anti-363 serum, the same number of X plaques is found by 
plating on K 12 (363)+ or on K 12/363. 

The results of this experiment show, therefore, that at 37° the X ma- 
terial carried in particles of phage 363 is able to multiply vegetatively 



347 



216 FRANgOIS JACOB 

if injected in X nonlysogenic bacteria. If it is assumed that, as in the case 
of bacterial recombination, most of the transduced X prophages develop 
when transduction is performed at high temperature, such an experi- 
ment enables one to estimate the fraction of particles of 363 carrying 
X material as between 10~® and 10~^ This proportion is of the same order 
of magnitude as the fraction of particles of 363 found to be able to 
transduce threonine or leucine characters. 

Prophage recombination. Prophage recombination between two geneti- 
cally different and complementary prophages has been shown to occur 
in transduction. In such experiments, phage X was used because its 
genetic system had been investigated extensively (Jacob and Wollman, 
19546; Kaiser, 1954). All the known markers of X are located on a single 
linkage group, which has been represented in Fig. 2. When sensitive 
bacteria are infected with genetically marked phages, some become lyso- 
genic. The phage particles released by such bacteria and their progeny 
remain genetically identical to the original particles used for infection. 
The markers available for the study of X genetics during the vegetative 
phase can therefore be used for studying the prophage. 

In the experiments described below, three markers have been used: 
ms (medium-sized plaques), Co (cocarde) and m, (minute plaques). 
These markers cover most of the known length of the X linkage group; 
in crosses involving these three markers the two parental and the six 
recombinant types can be distinguished easily and all are temperate. 
During the vegetative phase, the recombination frequency is about 7 
to 9 % between m^ and Co and 4 to 5 % between Co and mi. 

In the experiment reported in Table 6, lysogenic bacteria P678 
(XnisComf )/X were infected with particles of 363 grown on K12(Xm^Comi). 
The mixture was plated on galactose medium. After reisolation on galac- 
tose agar individual colonies were grown in broth and analyzed for the 
type of phage spontaneously released. It is seen in Table 6 that, of 240 
colonies, 213 released only the original type XmaCom^. The other 27 can 
be classified in two groups. In the first group, corresponding to 18 
colonies (7.5%), one or several of the original prophage markers have 

(P4) 
nrig rn5 g| s ce, c co mi 

3 1^ 3 L5 1.5 01 5 

Fig. 2. The linkage group of bacteriophage X. Figures indicate percentage of 
recombinants found in two factor crosses. 



348 



TRANSDUCTION OF LYSOGENY IN ESCHERICHIA COLI 



217 



TABLE 6 
X-Prophage Recombination through Transduction 



Number of clones 


Types of phage released 


Phage alleles present 


Phage alleles transduced 


213 


msCoinf 


niiComt 




2 


msComi 


miCorrii 


7n\ 


6 


msComi 


rriict nti 


ctm\ 


9 


nis CoiTii 


mtctm i 


mtctm; 


1 


msCoini'" 


mi.ctm-, 


ct 


1 


1 


'rnsComi^ 
msCoini 
msComi*" 


niiComt 


rrii 


3 


1 


msCoini 
msConii' 
msComi 


rriiComt 
ctrrii 


ctm, 




msComi*" mbContii 






4 


mtcoirii mtcomr 
msConii msComf 


niiComt 
mtctnii 


mtctmi 


1 


/maComr 
Ims'Comi'" 


niiCotnt 
mt 


nit 



Lysogenic acceptor bacteria P678(Xm5Comi^)+A were infected with phage 363 
grown on prototrophic lysogenic K12(Xm5'Comi)+. After plating on galactose 
medium, colonies were reisolated and tested for the type of phage released. 

been replaced by markers of the transduced prophage. In spite of the 
small numbers involved, it is clear that the markers of the prophage X 
are not independently substituted, the frequency of substitution de- 
creasing progressively from the m, to the m^ end of the X linkage group. 

In the second group, corresponding to 9 cases (3.8%), the bacteria 
were found to release, besides the original msComf type, one or more 
other types of X particles, even in single-burst experiments. This finding 
shows that alleles issued from the transduced prophages have been added 
to the original prophage. Here, again, the mi region is the most fre- 
quently added. 

These clones, in which bacteria carry two phage alleles at one or 
several loci, appear to be rather stable, since they were tested for phage 
production after two successive isolations on agar and growth in broth, 
corresponding to at least 50 generations. Nevertheless, single-burst ex- 
periments show that phage markers are very heterogeneously distributed 



349 



218 FRANCOIS JACOB 

in the population. During the multipHcation of such strains, bacterial 
clones segregate which have lost one or more prophage alleles. Very 
often, clones are found in which only the original maConii^ prophage re- 
mains. Other clones exhibit various types of phage allele combinations. 
Attempts to demonstrate some kind of diploidy for the Galh marker in 
these strains were unsuccessful. After two successive isolations on 
galactose-selective medium, colonies were grown in broth and streaked 
on EMB galactose agar. All contained only Gal^ bacteria. These results 
seem analogous to those found after mixed infection with two geneti- 
cally labeled X phages (Appleyard, 19546; Jacob, unpublished results). 
The results of this experiment show conclusively that the transduced 
prophage can either recombine with, or be added to, the prophage of the 
acceptor bacteria. Since the probability of being incorporated decreases 
from the mi to the ms end of the X linkage group and since, in addition, 
mi has no selective advantage over Ws in establishing lysogeny after in- 
fection, it seems likely (although the data are statistically not very sig- 
nificant) that the mi end of the X prophage is the closest to the Galh 
locus. Although no definite conclusion can be drawn from the small 
numbers of cases, the ratio of the probabihties of bringing either mi 
alone or mi and Co together (3 : 9) is not in disagreement with the ratio 
of the frequencies of recombination niiCo/coms found in crosses between 
vegetative phages (4:9). 

DISCUSSION 

The experiments reported in this paper show that the phage material 
present as a prophage in the donor bacterium can be transferred to an 
acceptor bacterium by means of a transducing phage. The transduced 
phage material can develop in the nonlysogenic acceptor bacterium when 
experiments are performed at 37°. It seems likely, however, that this 
phage material is transferred as a constituent of the donor bacterial 
chromosome, i.e., as a prophage, since in transduction, as well as in 
bacterial recombination, the same linkages between lysogeny and nutri- 
tional markers are found. Moreover, at 20°, the transfer of lysogeny oc- 
curs with about the same frequency as the transfer of nonlysogeny. 

In E. coli K12, two relative measures of linkage between two given 
characters, such as threonine and leucine, can be obtained, one by trans- 
duction and one by recombination. The relationship between these two 
scales of measure is about the same when lysogeny is one of the char- 
acters involved. In transduction experiments, lysogeny behaves there- 



350 



TRANSDUCTION OF LYSOGENY IN ESCHERICHIA COLI 219 

fore exactly like any other genetic character. These results confirm pre- 
vious findings of Lederberg and Lederberg (1953) and of Wollman (1953), 
gained by means of bacterial recombination and showing that lysogeny 
is controlled by a genetic determinant of the bacterium. In transduction 
experiments the same efficiency is found in the transfer of lysogeny to 
nonlysogenic bacteria and of prophage markers to already lysogenic 
bacteria, proof that the genetic determinant of the bacterium which 
controls lysogeny is the prophage itself. The three prophages which have 
been used in this study are likely to be located at closely linked loci. 
Their relationship is now the subject of more detailed investigation. 

Although we know little about the orientation of the prophage with 
respect to the bacterial chromosome, data concerning prophage re- 
combination during transduction seem to indicate that the m, end of the 
\ prophage is the closest to the Gah locus. If the size of the transducing 
piece is not too widely distributed and if recombination inside the pro- 
phage region behaves in the same way as in other regions, such an ex- 
periment may allow us to estimate the genetic size of the prophage as 
compared to that of the adjacent bacterial region. In the case of the ex- 
periment previously described and in spite of the small numbers in- 
volved, the mi-rris distance may be estimated at about 4 to 6% of the 
Galh-nii distance. Such an estimation is probably premature, because 
we still lack a definite picture of the relationship between the prophage 
and the bacterial chromosome. 

The case of bacteria found after transduction or mixed infection, which 
carry several pairs of prophage alleles that segregate during bacterial 
multiplication, is unclear. In such bacteria, the two prophages are likely 
to be located on the same region of the bacterial chromosome, since both 
additions and substitutions of prophage alleles are found in a single 
transduction experiment. 

One of the most striking facts derived from transduction of lysogeny 
is that a phage coat can contain, besides homologous genetic material, 
the genetic material of one or more other viruses. 

ACKNOWLEDGMENT 

The author wishes to thank Mile Y. Nicolas for valuable technical assistance 
and Dr. C. Levinthal for criticisms and help in the preparation of the manuscript. 

REFERENCES 

Appleyard, R. K. (1954a). Segregation of lambda lysogenicity during bacterial 
recombination in Escherichia coli K12. Genetics 39, 429-439. 



351 



220 FRANgOIS JACOB 

Appleyard, R. K. (19546). Segregation of new Ij'sogenic types during growth of a 
doubly lysogenic strain derived from Escherichia colt K12. Genetics 39, 440-452. 

Bertani, G., and Nice, S. J. (1954). Studies on lysogenesis. II. The effect of tem- 
perature on the lysogenization of Shigella dysenteriae with phage PI. J . Bacter- 
ial . 67, 202-209. 

Jacob, F., and Wollman, E. L. (1954a). Induction spontan^e du d^veloppement 
du bacteriophage X au cours de la recombinaison g^netique chez Escherichia 
coli K12. Comvt. rend. 239, 317-319. 

Jacob, F., and Wollman, E. L. (19546). Etude gtoetique d'un bacteriophage 
temper^ A' Escherichia coli. I. Le systeme g^n^tique du bacteriophage. Ann. 
inst. Pasteur 87, 653-673. 

Kaiser, A. D. (1954). A genetic analysis of bacteriophage lambda. Ph. D. Thesis, 
California Institute of Technology, Pasadena. 

Lederberg, E. H., and Lederberg, J. (1953). Genetic studies of lysogenicity in 
Escherichia coli. Genetics 38, 51-64. 

Lederberg, J. (1947). Gene recombination and linked segregation in Escherichia 
coli. Genetics 32, 505-525. 

Lennox, E. (1955). Transduction of linked characters of the host by bacterio- 
phage PI. Virology 1, 190-206. 

LwoFF, A. (1953). Lysogeny. Bacteriol. Revs. 17, 269-337. 

Stocker, B. A. D., ZiNDER, N. D., and Lederberg, J. (1953). Transduction of 
flagellar characters in Salmonella. J. Gen. Microbiol. 9, 410-433. 

Tatum, E. L., and Lederberg, J. (1947). Gene recombination in the bacterium 
Escherichia coli. J. Bacteriol. 53, 673-684. 

Weigle, J. J., and Delbrxjck, M. (1951). Mutual e.xclusion between an infecting 
phage and a carried phage. /. Bacteriol. 62, 301-318. 

Wollman, E. L. (1953). Sur le d^terminisme g^n^tique de la lysog^nie. Ann. inst. 
Pasteur 84, 281-293. 

Wollman, E. L., and Jacob, F. (1954). Lysogenie et recombinaison g^netique chez 
Escherichia coli K12. Compt. rend. 239, 455-456. 

ZiNDER, N. D. (1953). Infective heredity in bacteria. Cold Spring Harbor Sym- 
posia Quant. Biol. 18, 261-269. 

ZiNDER, N. D., and Lederberg, J. (1952). Genetic exchange in Salmonella. J. Bac- 
teriol. 64, 679-699. 



352 



Recombination between Related Temperate Bacterio- 
phages and the Genetic Control of Immunity and 
Prophage Localization' 

A. D. Kaiser and F. Jacob 

Department of Microbiology, Washington University School of Medicine, St. Louis, 
Missouri, and Service de Physiologic Microbienne, Institut Pastexir, Paris, France 

Accepted August 31, 1957 

Lysogenic bacteria are immune to infection with phage homologous to the 
prophage. By means of crosses between phages showing different immunity 
specificities it is shown that immunity specificity is determined by the same 
segment of the phage genetic material that controls the ability of the phage 
to lysogenize. Furthermore, this same segment, called the "ci region", also 
determines which locus is occupied by the prophage on the chromosome of 
Escherichia coli K12. 

INTRODUCTION 

Lysogenic bacteria possess and transmit to their offspring the power to 
produce bacteriophage. The phage that lysogenic bacteria produce is 
capable of making new lysogenic strains by infection of sensitive bacteria. 
By morphological, serological, and genetic criteria, the phage produced 
by such a lysogenized strain is identical with the phage used for the 
lysogenizing infection. Thus each lysogenic bacterium carries all the 
genetic information present in an infective phage particle. Nevertheless, 
disruption of lysogenic bacteria fails to reveal infective particles. The 
structure bearing the genetic information of the phage in a lysogenic 
bacterium is called a "prophage" (Lwoff, 1953). There is strong evidence 
that the prophage is the genetic material of the phage bound to a specific 
site of the bacterial chromosome (Lederberg and Lederberg, 1953; 
Wollman, 1953; Appleyard, 1953; Jacob, 1955). 

The presence of the prophage confers upon the host bacterium a re- 
sistance to infection with the homologous phage and its mutants. This 

' Part of this work was supported by grant E-1275 from the National Insti- 
tutes of Allergy and Infectious Diseases, National Institutes of Health, 
Bethesda, Maryland. 



Reprinted by permission of the authors and Academic Press, Inc. 
from Virology, 4 (3), 509-521 (1957). 



353 



510 KAISER AND JACOB 

property, called immunity, is phage specific since lysogenic bacteria re- 
main, with some exceptions, sensitive to all other phages that can infect 
their nonlysogenic counterparts. Immunity is reciprocal, in the sense 
that a bacterium lysogenic for any temperate mutant of a phage is 
immune against the wild type and its mutants. Several experiments 
(Jacob, 1954; Bertani, 1954; Goodgal, 1956) support the idea that the 
deoxyribonucleic acid (DNA) of homologous phage can penetrate into a 
lysogenic bacterium, suggesting that immunity is due to a block, which 
prevents phage multiplication inside the bacterium. 

Little is known about the mechanism of immunity or the origin of its 
high degree of specificity. It has been suggested that there is a relation 
between immunity and the specific location occupied by a prophage on 
the chromosome of its host (Lwoff, 1953). 

A new experimental approach to the problem of immunity was opened 
when a series of temperate phages, all active on E. coli K12, was isolated. 
Each of these phages has a different specificity of immunity: bacteria 
lysogenic for any one of them are still sensitive to the others. Yet, 
genetic recombination occurs between pairs of these phages when they 
multiply together in the same bacterium. Thus, the genetic control of 
immunity could be studied by phage crosses. Not only do these phages 
differ in their specificity of immunity, but they differ also in their lo- 
cation as prophages in the host. Each of them appears to occupy a 
specific location on the bacterial chromosome (Jacob and WoUman, 
1957). Phage crosses offer, therefore, the possibility of determining which 
part of the phage genetic material determines the specificity of im- 
munity and which part determines the specificity of the prophage lo- 
cation on the bacterial chromosome. The results of such crosses are the 
substance of this report. 

MATERIALS AND METHODS 

Only those materials and methods which were not described in a 
previous publication (Kaiser, 1957) will be given here. 

Media. EMB galactose agar contains 10 gm of Bacto-tryptone, 5 
gm of yeast extract, 2 gm of K2HPO4, 0.4 gm of eosin Y, 0.065 gm of 
methylene blue, 5 gm of D-galactose, and 15 of gm Bacto-agar per liter 
of distilled water. 

Bacteria. A galactose negative mutant of E. coli K12 strain C600 was 
selected by the penicillin method (Davis, 1948; Lederberg and Zinder, 
1948) following an irradiation with ultraviolet light. This mutant is 



354 



IMMUNITY AND PROPHAGE LOCALIZATION 511 

TABLE 1 

Immunity and Host Range of Phages X, 434, 82, and 21 











Host bacterium 








Phage 


















K12 


K12(X) 


K12(434) 


K12(82) 


K12(21) 


K12/X 


K 12/434 


X 


+ 


— 


+ 


+ 


+ 


_ 


+ 


434 


+ 


+ 


- 


+ 


+ 


+ 


- 


82 


+ 


+ 


+ 


- 


+ 


+ 


- 


21 


+ 


+ 


+ 


+ 


— 


- 


+ 



Key: + indicates ability of phage to grow on the host; — indicates inability 
to grow. 

Note: Loopfuls containing 2 X 10* or more phage particles were inoculated 
onto the surface of an agar plate seeded with bacteria. Lysis of the bacteria within 
the inoculated area was taken as the sign of phage multiplication. 

related to the gali~ of Morse et al. (1956), since it can be transduced to 
gal+ by XHFT gala" but not by XHFT galr. 

Phages. Coli phages 434, 82, and 21 (Jacob and Wollman, 1956) are 
capable of lysogenizing E. coli K 12. All three are serologically related to 
phage X. Phage X is the wild type described previously (Kaiser, 1957). 
The ability of phages X, 434, 82, and 21 to grow on various lysogenic and 
resistant strains of E. coli K12 is shown in Table 1. It may be seen that 
each phage is distinguished from the others by its ability to grow on all 
of the lysogenic strains except the homologous one: each of the four 
phages has a different immunity specificity. 

Crosses. The technique of crossing these phages is the same as the one 
already described for XXX crosses (Kaiser, 1957). 

Because many stocks of X mutants are available, crosses were made 
between X and each of the other phages. 

RESULTS 

1. Genetic recombination between X, 4^4, 82, and 21. The first series of 
crosses involved phages 434, 82, or 21 as one parent, and X marked either 
with ms, C02, Ci, coi, or mi as the other parent. A map giving the position 
of the markers on the X linkage group is given in Fig. 1. The purpose of 
these crosses is to see if 434, 82, and 21 carry and are able to transfer 
to X the wild type alleles m5+, co^^, Ci+, coi+, and wt+. Thus, for example, 
Xms, which forms medium-sized plaques, was crossed to 434 and the 
offspring of the cross plated on K12(434), to examine only the phages 
with the immunity specificity of X. The presence of some large plaques, 



355 



512 KAISER AND JACOB 



-I H- 



, C02C|C0| mi 1% 

\ RECOMBINANTS 



1^ I I I 1 II I I I \-\^ 

47__30l5952cp, 

I n 

Fig. 1. A map representing recombination frequencies observed between pairs 
of X mutants (Kaiser, 1957). The c segment is shown on an expanded scale. 

m5+, indicated that m^^ was present in 434 and that recombination can 
occur between the m^ locus and the locus controlling immunity specificity. 
Similar experiments were carried out with C02, Ci, coi, and mi. The markers 
CO2, Ci, and coi cause the production of clear plaques, their corresponding 
wild alleles controlling the production of turbid plaques. The marker 
mi causes minute plaques with a halo. The results of these crosses, sum- 
marized in Table 2, are that in the cross X X 434 recombination can occur 
between ms, CO2, coi, and mi on the one hand, and the loci controlling the 
specificity of immunity on the other hand. In crosses X X 21, recombina- 
tion of ms, C02, and mi with immunity specificity can occur and in crosses 
X X 82 the loci ms, coi, and mi can recombine with the loci controlling 
the specificity of immunity. The pattern of recombination is similar in 
the three types of crosses of X with 434, 82, and 21 in that recombination 
can occur at the ends, m^ and mi, but not in the middle, Ci, of the X 
linkage map. The three types of crosses differ, however, in the length 
of the middle region which fails to recombine with the loci controlling 
immunity specificity. In the cross X X 21 this region includes both C\ 
and coi, in the cross X X 82 both coi and Ci, while in the cross X X 434 
only cx fails to recombine. Because the region within which recombina- 
tion fails to occur is the shortest for the cross X X 434, a more detailed 
study of recombination between X and 434 was undertaken. Crosses 
were made between 434 and a selected series of clear mutants of X. 
The clear mutants were selected to mark different regions of the c 
segment. 

The c segment of X consists of at least three regions. Each region is 
probably responsible for the performance of a specific function in lyso- 
genization (Kaiser, 1957). One mutant, number C44, was selected from 
region III ; 4 mutants, C47, C30, Ci, and C50, from region I ; and two mutants. 



356 



IMMUNITY AND PROPHAGE LOCALIZATION 513 

TABLE 2 

Occurrence of Wild Type Recombinants with the Immune Specificity ok 
X IN Crosses between a Series of X Mutants and Three Related Phages 



Ci 


■OSS 








N Mutant employed 








nib 


COi 


C\ 


COl 


mi 


434 


X 


X 


+ 


+ 


— 


+ 


+ 


21 


X 


X 


+ 


+ 


- 


- 


+ 


82 


X 


X 


+ 


— 


— 


+" 


+ 



" Recombinants arose which were more turbid than Xcoi, but less turbid than 
Xcoi"*". 

Key: + signifies the occurrence of wild type recombinants with the immune 
specificity of X; — indicates failure to detect wild type recombinants among 1000 
offspring with the immune specificity of X. 

Note: Several mutants of X (X rui, X cOi, X Ci, X coi, andX mi) were crossed to 434 
and the phage progeny was plated on K12(434), where plaques with the wild 
type morphology and the immunity specificity of X could be detected. Analogous 
crosses were carried out with phages 21 and 82. 

C42, and COl, from region II. A map showing the positions of the clear 
mutants employed is given in Fig. 1. 

The crosses made were 434 X XCi, where Cx represents individual 
members of the selected series of clear mutants. The progeny of each 
cross were plated on K12(434) and on K12(X). The presence of turbid 
plaques on K12(434) demonstrates the presence in 434 of a Cx^ allele, 
which can recombine with the locus controlling immunity specificity of 
X. Clear plaques on K12(X) reveal phages with the immunity of 434 
which have picked up a Cx allele from X. A comparison of the proportion 
of turbid plaques on K12(434) with the proportion of clear plaques on 
Kr2(X) serves as an internal control, because they are reciprocal re- 
combinants and should therefore have the same frequency. When 434 
alone is plated on K12(X), there are 0.07 % clear plaques due to mutants 
in the 434 stock. This proportion is, naturally, present as a background 
in the yield of all of the crosses. The proportion of turbid mutants in a 
stock of clear X is less than 1 in 10* and is therefore too low to be de- 
tected in the crosses. 

The results of this series of crosses, given in Table 3, show that 
recombinants arise from the crosses with mutants Cu, C42, and coi, but 
for none of the others. There is, therefore, recombination between 434 
and X in regions II and III of the c segment but not in region I. The 



357 



514 



KAISER AND JACOB 



TABLE 3 
Crosses X Clear X 434 Turbid 





Plated on C600(434) 


Plated on C600(X) 


Plated on 


C600 


Cross 


Num- 
ber 
turbid 


Number 
clear 


% Turbid 


Num- 
ber 
turbid 


Num- 
ber 
clear 


% 

clear 


Nun- 

ber 

turbid 


Num- 
ber 
clear 


% Clear 


\cu X 434 


81 


2232 


3.5 


2977 


119 


3.8 








Xc-47 X 434 





12528 


<0.01 


8166 


4 


0.05 








Xcto X 434 





3080 


<0.03 


3632 


2 


0.06 








Xfi X 434 





9132 


<0.01 


5518 


4 


0.07 








Xcso X 434 





5212 


<0.02 


3615 


1 


0.03 








XC42 X 434 


41 


7612 


0.54 


6542 


32 


0.49 








Xcoi X 434 


65 


5182 


1.2 


4249 


170 


3.8 








Xc+ X 434 














1550 





<0.1 


434 alone 








3000 


1 


0.03 


2500 


3 


0.1 


Xcu alone 


2 


3744 


0.05 














Xc42 alone 





4296 


<0.03 














Xcoi alone 





2502 


<0.04 















Note: Progeny of crosses of the type X clear by 434 turbid, plated on C600, 
C600(X), or C600(434) were scored for the number of clear and turbid plaques. 
The clear mutants of X employed in these crosses were selected to mark different 
regions of the c segment: C47, C30, Ci, and C50 to mark region I, C42 and coi to mark 
region II, and C44 to mark region III. 

limits of recombination between X and 434 lie, on one side, between 
C44 and C47 and, on the other side, between C50 and C42. The precise location 
of the limits between these pairs of markers is suggested by the fre- 
quencies of recombination with C44 and C42, i.e., 3.5% recombination to 
the right of 44 and 0.54 % recombination to the left of C42. 

The recombination pattern in X X 434 crosses can be understood if it 
is postulated that the Ci region controls immunity specificity, that there 
is a Ci region specific for the immunity of X and a Ci region specific for 
the immunity of 434. 

2. 434-^ Hybrid. Hybrid phages which have only the immunity de- 
termining region of 434 and the rest of their chromosome from X were 
isolated from 5 successive backcrosses of 434 to X. The first step was to 
cross 434, which forms tiny turbid plaques, with Xci, which forms large 
clear plaques. A few large turbid plaques were found among the progeny. 
They were tested and all found to have the immunity of 434. The second 
step was to cross one of the large turbid 434 to XC44 and to isolate a 434 C44 



358 



IMMUNITY AND PROPHAGE LOCALIZATION 



515 



by plating on K12(X). Step three consisted in replacing Ca by C44+ by 
crossing to Xci and picking the turbid recombinants on K12. In step 
four, 434 C44+ was crossed to XC42. A 434 C44"'"C42 was selected by picking 
a clear plaque on K12(X). The final step consisted in replacing C42 by 
Ci2^ by crossing 434 C44+C42 to Xci and plating on K12. At each stage in the 
backcrosses the immunity pattern was checked. This sequence of 5 
successive backcrosses would be expected to replace all of the parts of 
the 434 chromosome capable of recombination with the homologous 
parts from X. We have, therefore, a phage which should possess the Cj 
region of 434 embedded in an otherwise X genome. This phage will now 
be designated 434hy. 

The hybrid character of 434hy is evidenced by the following properties. 
Apart from its immunity pattern, which is that of 434, phage 434hy 
behaves like X. On K12 it forms large turbid plaques identical to those of 
X and it has the host range of X. Whereas phage 434 is inactivated by 
anti-X serum at a slower rate than is X, the hybrid is inactivated at the 
rate characteristic of X, as is shown in Fig. 2. Further evidence for the 
homology of 434hy with X is shown by a comparison of X X 434hy with 
XXX crosses. In a cross XmsCi X 434hy mi, 12 % recombination was ob- 
served between ms and Ci and 4 % between Ci and mi. A corresponding 
cross between Xm^c and Xmi gave 8% and 5% recombination, respec- 



10" 



10' 



\ 0^ 


~~"^v^34 


\ D 


^\P 


\s 


O-x 
D- 434Hy 


Minutes — 
5 10 


15 20 



Fig. 2. Neutralization was measured by incubating a mixture of phage and a 
1/100 dilution of rabbit anti-X serum at 37°. At the times indicated aliquots were 
withdrawn, diluted at least 1/10'' and plated on C600. The surviving fraction is 
the ratio of the number of plaques at time t to the number at time zero. 



359 



516 KAISER AND JACOB 

tively, in the two regions. The absence of a significant difference in 
recombination frequencies suggests that 434hy is homologous to X. 
Since the Ci region is only about 1 % long, absence of homology there 
would not be expected to depress recombination frecjuencies significantly 
in this cross. 

Bacteria lysogenic for 434hy behave exactly like K12(434) in response 
to infection by X, 434, Xc, 434 c, and434hy. Thus the 434hy prophage 
determines the immunity pattern characteristic of the wild type 434. 

The existence of hybrids with a Ci region from 434 and the rest of their 
chromosome from X, which have the specific immunity pattern of 434 
but not that of X, demonstrates that the Ci region (or a portion of it) 
controls the specificity of immunity. 

Wild type T4 is capable of multiplication on K12(X), but the rn 
mutants of T4 are not (Benzer, 1955). T4rii was tested on K12(434) and 
on K12(434hy); it was found to multiply on both strains. T4r+ was 
able to multiply on both also. This suggests that the block in T4rii 
multiplication is caused by the Ci region of X since replacement of the 
Ci region of X by the Ci region of 434 eliminates the block. 

3. Chromosomal localization of 434hy prophage in E. coli K12. Prophage 
X and prophage 434 occupy distinct but closely linked loci on the linkage 
map of E. coli K12 near certain galactose loci (Jacob and Wollman, 
1957). It must, therefore, be determined whether prophage 434hy oc- 
cupies the X locus, the 434 locus, or neither. 

In K12(X) prophage X occupies the X locus and in K12(434) prophage 
434 occupies the 434 locus, by definition. Allelism tests between 
K12(434hy) and K12(X), and between K12(434hy) and K12(434), 
should localize the 434hy prophage. 

Allelism tests were performed by means of transduction with phage 
PI (Lennox, 1955). The occurrence of transductants nonlysogenic for 
the prophage of the donor and the prophage of the recipient would 
indicate that the two prophages are nonallelic. The absence of non- 
lysogenic transductants indicates allelism, subject of course to the limita- 
tion in the number of transductants examined. 

In the experiments to be reported, a stock of Pike (Lennox, 1955) was 
prepared on a gal+ donor strain. The gal~ recipient was infected with 
this Pike stock and gal+ transductants isolated, purified, and tested for 
prophage by replica plating on C600, C600(X), and C600(434). The 
recipient strains had all been rendered /X/434 to prevent infection by 
free phage X or 434. 



360 



IMMUNITY AND PROPHAGE LOCALIZATION 



517 



TABLE 4 
Prophage Segregation in Transduction 





1^ 


Recipient gal" 




Transductants 


gal* 






Donor ga 


Non- 
lysogenic _ 


Lysogenic 

Produce phage with the 
immunity specificity of 


Total 




X 


434 


X and 434 




(434hy) 


A 


(X) 


75 


267 


39 




16 


397 


(434hy) 


B 


(X) 


103 


122 


114 




59 


398 


(434hy) 




(434) 








789 







789 


(434) 




(434hy) 








785 







785 


(X) 




(434hy) 


36 


38 


320 




21 


415 


(434) 


A 


(X) 


39 


228 


25 




6 


298 


(434) 


B 


(X) 


23 


370 


7 







400 


(X) 




(434) 


18 


29 


71 




2 


120 


(X) 




(X) 





943 










943 


(434) 




(434) 








200 







200 


(434hy) 




(434hy) 








400 







400 



Note: The transductions were performed by infecting the recipient with phage 
PI that had been grown on the donor. The infected mixture was incubated 1 
hour at 25°, then spread on EMB galactose agar. Galactose positive transductants 
were isolated, purified bj^ two serial single-colony isolations and then scored for 
the type of phage production by replica plating onto K12, K12(X), and K12(434). 
A and B indicate strains isolated independently using the same stock of phage 
and sensitive bacterial strain. The donor strains were all prepared from C600 
and the recipient strains from C600 gal~. 

The results of these tests, presented in Table 4, are that nonlysogenic 
transductants did arise when the pair K12(X), K12(434hy) was tested, 
independently of which was the donor and which the recipient. How- 
ever, no nonlysogenic transductants arose when the pair K12(434), 
K12(434hy) was tested. Therefore, prophage 434 hybrid occupies the 
434 locus. Incidentally, the absence of nonlysogenic transductants when 
the pairs K12(X), K12(X) or K12(434), K12(434) were tested is a strong 
argument in favor of the uniqueness of the X and the 434 loci. 

If prophage 434hy occupied the same locus as prophage 434 in K12, 
then the same frequencies of recombination would be expected between 
the locus of prophage 434hy and the locus of prophage X as between the 
locus of prophage 434 and the locus of prophage X. Referring to Table 4, 
it may be seen that when the donor was lysogenic for 434hy and the 



361 



518 KAISER AND JACOB 

recipient lysogenic for X an average of 22 % (75 + 103/397 + 398) of the 
gal+ transductants were nonlysogenic. But when the donor was lysogenic 
for 434 and the recipient lysogenic for \, 9% (39 + 23/298 + 400) of 
the gal+ were nonlysogenic. In the reciprocal transductions where the 
donor was lysogenic for X and the recipient lysogenic for 434hy or for 
434, the frequencies of nonlysogenic strains among the gal+ were 9% 
(36/415) and 15% (18/120), respectively. The significance of the differ- 
ences in each case is questionable for two reasons. First, differences of 
the same order were observed when the same transduction was repeated 
with independent isolates of the same donor strain, as may be seen by 
comparing examples A and B in Table 4. Second, the hypothesis that 
prophage 434hy and prophage 434 occupy different loci would require 
that a particular order be assignable to the loci of the three prophages 
involved. Neither the order X ~ 434 — 434hy nor the order X — 434hy — 
434 is consistent with the frequencies of nonlysogenic transductants ob- 
served in reciprocal transductions. For, whereas the order X — 434 — 
434hy might have been indicated when the donor was lysogenic for 434 
or for 434hy and the recipient lysogenic for X due to a greater frequency 
of nonlysogenic strains in the latter transduction than in the former, 
the reverse order X — 434hy — 434 would have been indicated by the 
reciprocal transductions when the donor was lysogenic for X, due to a 
greater frequency of nonlysogenic transductants when the recipient was 
lysogenic for 434 than when the recipient was lysogenic for 434hy. 
These data, therefore, are not incompatible with the idea that prophage 
434hy occupies the same locus as prophage 434. 

Allelism of prophage 434hy to prophage 434 was independently con- 
firmed by the bacterial cross Hfr T+L+S^ gal+(434) X F+T-L-S^(434hy) 
gal~/X/434. Among 400 gal+S'' prototrophs no nonlysogenic strains were 
found. 

Bacteria lysogenic for both X and 434 or for X and 434hy were found in 
some of the transduction experiments reported in Table 4. Ten strains of 
each type were examined in detail ; all of them had the following proper- 
ties. They produced two types of phage, one type which plated on K12(X) 
and another type which plated on K12(434). When induced with ultra- 
violet light and plated, before lysis, on C600, they produced mottled 
plaques like those of C600 mixedly infected with X and 434. These strains 
were immune both to X and to 434 and did not support the growth of 
T4r„. 



362 



IMMUNITY AND PROPHAGE LOCALIZATION 519 

DISCUSSION 

In crosses between X and any one of the other phages studied, 434, 82, 
and 21, genetic recombination observed in the ms and mi regions indi- 
cates clearly that genetic homology exists between the linkage maps of 
these phages, at least in their outer parts. That the degree of homology 
may vary according to the pair of phages is shown by the fact that in 
crosses with X, the regions of the map within which recombination can 
occur is different for 434, 82, and 21. 

The most interesting of these three phages is 434, which appears to 
exhibit complete homology with X except for a small region of its linkage 
map, the Cj region. The Ci region of the genetic material of phage X has 
been defined as that segment of the linkage map within which are lo- 
cated all mutations (22 out of 22 studied) that suppress the ability of the 
phage to lysogenize or, at least, reduce the lysogenization frequency to 
less than 10"^ (Kaiser, 1957). Mixed infection with different pairs of 
Ci mutants have indicated that the Ci region is probably a single cistron 
or functional unit as defined by Benzer (1957). 

The failure to find intra-Ci recombinants in a X X 434 cross may be 
explained by two kinds of hypotheses. Either the Ci regions of X and 434 
fail to pair, or recombination occurs but the recombinants are inviable 
because of the hybrid character of their Ci region. In both models, struc- 
tural dissimilarity of the Ci regions of X and 434 would seem to be implied 
by the absence of recombination there. 

The main result obtained by comparing the properties of X and 434hy 
is that these two phages, which have in common the whole linkage map 
except the Ci region, differ both by their immunity pattern and by their 
specific location as prophages on the bacterial chromosome. The conclu- 
sion, therefore, seems inescapable that the C\ region controls not only a 
reaction involved in the lysogenization process, but also the immunity 
pattern and the specificity of prophage location. The question arises, 
therefore, whether these properties are but different expressions of the 
same function. 

To understand what this function might be it is necessary to consider 
the nature of prophage. Prophage is defined as that element in a lysogenic 
bacterium which is responsible for the production of a particular bac- 
teriophage (Lwoff, 1953). Bacterial crosses between K12 strains lysogenic 
for different mutants of X (Appleyard, 1953) as well as transduction of 
prophage (Lennox, 1955; Jacob, 1955) demonstrate that prophage is the 



363 



520 KAISER AND JACOB 

genetic material of a phage associated witii a particular locus of the 
bacterial chromosome. If then the Ci region of the phage were responsible 
for the association, the three functions — lysogenization, localization, 
and immunity — would all depend on the Ci region. In the following 
discussion this idea will be explored in more detail. 

The simplest way to explain how the Ci controls the specificity of 
prophage localization would seem to be that the prophage is attached to 
the bacterial chromosome by its Ci region. The Ci region of X would attach 
to the X locus in K12, the Ci region of 434 to the 434 locus. A slight 
structural change in a Ci region might prevent attachment. Thus it would 
be easy to see how a mutation in the Ci region could prevent 
lysogenization. 

The role of the Ci region in immunity may be summarized in the fol- 
lowing way. The fact that X but not 434hy can reproduce in K12(434) 
shows that immunity is directed against the Ci region of a superinfecting 
phage. The fact that X but not 434 can reproduce in K12(434hy) shows 
that immunity is governed by the Ci region of the prophage. Thus the 
prophage controls the inhibition of vegetative multiplication of superin- 
fecting particles possessing an homologous Ci region. This may be ex- 
plained by two types of hypothesis. Either immunity operates in the 
bacterium at the chromosomal level and is due to some steric hindrance, 
or immunity is a cytoplasmic expression of a genetic function of the pro- 
phage. This could be either the production of an "immunity substance" 
(Bertani, 1955) by the prophage-bacterial chromosome system or the 
inhibition by the prophage of the production by the bacterium of a 
substance necessary for the multiplication of homologous particles. No 
decision between these hypotheses can be reached until further attempts 
have been made to dissociate the genetic control of lysogenization, 
specificity of prophage location, and specificity of immunity. 

The Ci region of the prophage also controls the ability of T4rii to 
multiply on lysogenic K12. The mechanism of this effect remains un- 
known but the same hypothesis as those proposed for immunity may be 
considered (Lederberg, 1957). 

ACKNOWLEDGMENTS 

We want to thank Dr. Andre Lwoff in whose laboratory at the Institut Pasteur 
much of this work was carried out. During this time the senior author was a fellow 
in cancer research of the American Cancer Society. 



364 



IMMUNITY AND PROPHAGE LOCALIZATION 521 

REFERENCES 

Appleyard, R. K. (1953). Segregation of lambda lysogenicity during bacterial 

recombination in E. colt K12. Cold Spring Harbor Symposia Quant. Biol. 18, 

95-97. 
Benzer, S. (1955). Fine structure of a genetic region in bacteriophage. Proc. 

Natl. Acad. Sci. U.S. 41, 344-354. 
Benzer, S. (1957). The elementary units of heredity. In The Chemical Basis of 

Heredity (W. D. McElroy and B. Glass, eds.), p. 70, The Johns Hopkins 

Press, Baltimore. 
Bbrtani, G. (1954). Studies on lysogenesis. III. Superinfection of lysogenic 

Shigella dysenteriae with temperate mutants of the carried phage. J. Bacteriol. 

67, 696-707. 
Bertani, G. (1955). The role of phage in bacterial genetics. Brookhaven Symposia 

in Biol. No. 8, 50-56. 
Davis, B. D. (1948). Isolation of biochemically deficient mutants of bacteria by 

penicillin. J. Am. Chem. Soc. 70, 4267. 
GooDGAL, S. H. (1956). Incorporation of phage DNA on lysogenization of Shigella 

dysenteriae. Biochim. et Biophys. Acta 19, 333-340. 
Jacob, F. (1954). Les Bacteries Lysoghnes et la Notion de Provirus. Monographies 

de rinstitut Pasteur, Paris. 
Jacob, F. (1955). Transduction of lysogeny in Escherichia coli. Virology 1, 207- 

220. 
Jacob, F., and Wollman, E. (1956). Sur le processus de conjugaison et de recom- 

binaison chez Escherichia coli. I. L'induction par conjugaison ou induction zy- 

gotique. Ann. inst. Pasteur 91, 486-510. 
Jacob, F., and Wollman, E. (1957). Genetic aspects of lysogeny. In The Chemical 

Basis of Heredity (W. D. McElroy and B. Glass, eds.), p. 468, The Johns Hop- 
kins Press, Baltimore. 
Kaiser, A. D. (1957). Mutatio